Afternoon. Thank you very much for joining us today. We have a reasonably full house today. Thank you very much. Before I proceed, I'd like you to know that all my problems have been solved during the lunch. Between KWAP, [Uncertain], and PNB, they've given me all the answers what to do. Is that correct? So I just gave them a couple of hypothetical questions, and interesting. We had a good, interesting debate. Like, shall we list Celcom, for example, right? Or shall we delist XL? So that's a good question. Normally, you ask me the questions, and all rumors have been created about us. This time, I'm going to create rumors for a change. A nyway, thank you again. Y ou must be wondering why we focus on excellence.
So to put into context the whole of half day today, we're not going to spend. Of course, you can ask, for sure. In fact, we have time for Q&A. We're going to spend so much time on so many other strategies that we have put in place because, in essence, nothing has dramatically changed. I will actually talk about it a bit. But we're going to zoom in 80% of the time on Operational Excellence. Now, to put into context, if you look at our 12-year or 11-year history, every company has done either very well or extremely well when it comes to revenue growth and market share growth. Every operation's increased market share is some marginally 2-3%, some significantly. EBITDA growth is also. If you look at the chart, EBITDA growth is spectacular. Pretty good. That can be better, but quite good.
But our profit growth has not come in correspondingly. S tarting from end of last year, we presented, if you remember last year, we talked about shifting gear, right? We talked about, okay, not that we don't want to grow revenue anymore, but we want to grow revenue profitably, but we want to focus on what it takes to show higher, better EBITDA, better margin, and most of all, better profit and cash. T hat's the focus now. Now, this is a bit dangerous because if we do it wrongly, we will be at expense of our future and our revenue growth. But we think we can do still very well on revenue. We might sacrifice a bit of revenue growth if it's not profitable, but we will focus a lot on EBITDA and profit and cash.
Given that theme that we started last year on shifting gear, we want to focus on Operational Excellence. W e will define Operational Excellence later on. You'll see it's not necessarily equal to cost. There are a lot of things. It's running the company efficiently, and obviously, the byproduct will be profit in cash, but we will still manage, we will manage to have a strong revenue growth at the same time. T hat's the reason behind the theme. If I can get this to work. Can I? Where should I be pointing this to? Okay. So all of you have seen this. Back in 2015 and early 2016, we talked about our new ambition, right? Previous to that is a regional champion, right? W e said we want to be next generation digital champion.
I know it means so many things to so many people, but basically, we don't want to be called a telco by 2022. We want to evolve into a different kind of company and in terms of products, services, the way we operate, and so on. It's an extremely ambitious goal, but we think we can do that. But to do that, we are building into the three blocks. How do we convert or transform all our so-called traditional mobile operators to converged digital operators? How do we focus on our digital businesses from the 23 at one point in time to three, but big three ones that one or two or three can become unicorns? And how do we convert or transform our Edotco business to become a top five in the world? T hat's why we said in 2015 and 2016.
In 2017, we told you that there are eight things we are focusing on. Number one and the most important one is performance, performance, performance. But we also want to look at new growth areas. We want to improve every aspect of our IT and network, especially. We want to be clear into how we differentiate ourselves from a branding and actual customer value proposition and industry, how we want to rationalize the industry into the three-player market in most markets, and how do we look at ourselves monetize rightfully in areas that we need to monetize to eliminate the value of our data. Of course, stakeholder management and sustainability, very important. And lastly, how we make changes organizationally. W e talked about that back in, we launched this in late 2016, but in 2017, we were working on that.
Now, of course, you all know very well that the number five industry rationalization, we tried to merge with Telenor for whatever reason didn't quite happen. But we are still looking at a few possibilities as we speak today. Nothing imminent, I must say, but we are still exploring that. A t the end of 2018, more or less last year, same time, we said that, you know what, we're going to shift gear. This is basically what I've said, that from very much focus on growth and revenue and all that, which has still been the case, but we shift gear a bit to focus on profit and cash. T his is what we said the last time around. T hat's basically what we said, the evolution of our strategy or the tactical adjustment that we've made over the years. Okay. So you heard the third-quarter results.
Not perfect, almost perfect so revenue growth, almost everybody grew double-digit growth. EBITDA, almost everybody grew EBITDA, except for himself. Smart, same thing, and free cash flow, same thing. You will get the actual numbers later on. You have seen yourself, but Vivek will show the actual numbers but just to give you a flavor that when we say we want to do it, we are so far, so good, very much on track. Now, the one with a lot of rates, the Ncell, it's a pity that as we launched this, you probably, I'm sure you've heard that the regulator imposed a tax on data. That's quite hurtful for us in a country where the spending capability is quite limited so just in case you're not aware of, they increased what we call the TSC, Telecommunication Service Charge, I think, from 0% to 13% for data, from 0% to 13%.
That's very hurtful because now, instead of getting MYR 100 for the equivalent or 100, whatever the denomination, the customer only gets 87. Or put it the other way around, to get MYR 100, they have to pay 113%. It is bad enough for any country, but for a country like Nepal, it's extremely hurtful. So it unfortunately hurt us quite badly. Not to mention that ILD, which is not, that part is nothing surprising. We knew it, and we factored for it already when we acquired the company. That's gone down also quite a lot. Almost more or less, maybe slightly worse than we thought, but that's not the main reason. The main reason is the core business itself has been affected. Now, we believe that it will be readjusted by itself next year, but this year, the impact will be negative.
D espite doing all the above, focus on profit and cash and PATAMI, we perform number one, number two in all the markets in almost all the areas. Again, to remind ourselves and all of you that we are not forgoing revenue share or market share completely. Okay. T o go down a bit further for the Q on Q performance, you can see everybody improved their ARPU. Dialog reduction also because of tax reasons also. You probably have heard. Sorry, the floor price. They removed the floor price, so there's a bit of price fall last year. This year, we hope to improve next year and Ncell for the reason that we mentioned. That's ARPU. And m argin, everybody improved except Ncell. Quarter on quarter. F inally, EBITDA margin, everyone has improved, right?
In this case, we look at it all the way from first quarter 2018, first Q 2018 last year to third quarter this year. You can see the trend. Everyone trended upwards. So again, what we're trying to show is that when we say we want to do it, we mean it, and we have executed it so far, so good. Now, these are the five building blocks of Operational Excellence. Now, you'll be hearing a lot from us, the word Operational Excellence, many, many times. We just want to make sure that it's being really understood what we're trying to do. There are five components of Operational Excellence from Axiata perspective, right? It's not equal to cost only. It's much more than cost. The number one, of course, we still want to grow, but profitably. There's an element of short-term and medium-term and long-term.
We will skew it initially towards short-term and medium-term, right? W e want to show any growth to be profitable. In fact, we just completed our business plan for next year, although we can't reveal yet. We will announce in January or so, or February rather. You will see that we will still be growing, and we make what we have asked all the operating companies to do is take out all those long payback investments from our business case, from their business plan, and then we pick. So we come up with a baseline business plan. We will show very good, very healthy improvement in profit and cash, and those that have a long payback or impact that are, of course, very poor IRR and long-term payback, of course, easily rejected.
However, there's a bucket of investment cases whereby it is very good IRR, however, bad impact to profit the next one year or two. W e separate them for the first time. We separate them from all the business plans from the operating companies. W hat we did during the two-day session with the Axiata Board, we did some what we call capital allocation. In other words, not all we approved because we're trying to balance long-term and short-term. We approved many, but not all. So net-net, we will still show a good profit growth. That's what we want to achieve without impacting our long-term. It's a tough balancing act. Some of them are pretty good IRR, but severe P&L issue if you do that, right? So that's rejected. S ome will be good IRR, very good IRR.
Maybe the payback can be three, four years, but maybe if we don't do it, it will affect our future. So we might still proceed with that. I t's an extremely conscious and very disciplined process, but Operational Excellence is first about profitable growth, not only in general. For example, we will still be expanding selectively to Ex-Java in XL, for example. We will go wide, but we will be very selective in areas that we know that we're going to make it. Generally, we are number two there, the only two players there. We will still invest in home broadband, but we will reduce the original thought of fiber to the homes, but we will substitute that with fixed wireless access. And therefore, we might invest in spectrum instead, which can be used for fixed wireless access and the normal mobile, for example.
We will invest heavily and try to achieve major successes in enterprise where the payback is a lot shorter. That will be invested so that we can have growth in this area, right? T hat's the first point I want to note. The second point is this is a bit tricky. Many times when we look at how to reduce costs, we think about this is a scenario. I want to reduce cost of network rollout, cost of additional sites, cost of operating network, cost of IT, cost, and so on and so forth. But we say to ourselves, you know what? Before you even talk about reducing costs, let's challenge our—what do you call—what do you mean by customer satisfaction? What do we mean by service level?
Now, we will not be focusing as much, but just to give you a flavor, there are many things we went through as anally as detailed as possible. We'll find that many things you don't really have to do, and customers don't even realize the difference. You don't even realize our favorite, of course, I mentioned last year, YouTube, right? If I give you—I can show you on the screen 360 pixels or 480 versus 4K. On a small screen, you can't even tell unless some of you in this room have tremendous sharp eyes or whatever you call it. For 95% of us, you can't even tell the difference. T hat's just a simple—that's a very simple example. At least you might think.
But if you think that more than 50% of our traffic to 70% of our traffic is actually video, then you know what I mean by that. We are talking about a few hundred million savings just by simple things like that. I can give a long list. Even, for example, generators. We have some of the companies have generators because we are a very much engineering company. We want to make sure that we always have generators ready to back up. But frankly, we don't need many of them. We can deploy them in maybe half an hour later. It's not going to end the world. The other thing is, in fact, luckily we have EDOTCO in our operating company. Most of our service level from EDOTCO perspective is similar, almost clustered by customers side by side. We are challenging that. You know what?
I don't need to be everything the same. In some remote areas or some areas, our average is 99.74%. 99% is good enough. In some areas, it's got to be 99%. In some areas, it's got to be 95%. So we do cluster by cluster. You will see the word fragmentation by one, which has been used in marketing, will be used in all those operations. By site, depending on the application, depending on the device, depending on the customer, depending on the time of the day, you will have different service levels. Now, it's going to be extremely extensive across the board. But the point here is that we can achieve customer satisfaction without gold plating everything, right? Sounds cliché to some extent, but frankly, it's not. If you go down to the detail, it's not, right? And of course, cost itself, right? That is the crux of it.
But again, as the major part of Operational Excellence, but not the only. And then digitization, you will see that we are. I remember I mentioned to you last year that we started digitization probably among the earliest, not just industry, across all industries. We were struggling. What is that word called? I remember I even asked the McKinsey and the BCG, "Tell me, what is this thing called?" And we all thought, "Yeah, maybe digitization, maybe digitalization." We won't talk about the words, about the exact word. But unfortunately, we started it, but we didn't push enough. We started in 2012 or so, 2013, 2014, by 2015, and 2016, we were behind most of our competitors, which we believe in. Except for, I mean, to be fair, except for two or three of our operations. But the rest, we were behind. We were behind.
In 2016 and 2017, especially, we said, "You know what? We have to make this group-wide concerted structured effort." And we put KPIs in all the countries. Fast forward, by 2018 and especially now, we are either equal or better in most operations. And then we asked the next question, "Can this deliver tangible cost reduction?" The answer, absolutely yes. Again, that will be another presentation afterward by Dr. Hans, who will show you. We are trying to measure it. To be frank, it's not so easy to measure, but we're trying our best. Of course, analytics is different, but it's part of it too, right? How we analytically looking at not just from a revenue perspective, but from a cost perspective. It ties back to number two, right? If we have strong analytical AI and proper analytics model, we can reduce cost with minimal impact to customer.
That's the idea. And last but not least, about people, organization, process, which is pretty obvious. And the culture. I mean, to be frank, our culture for the last is a macho culture. We gain market share. We are very happy about it, but not good enough when it comes to cost, right? Maybe I wouldn't say it's bad, but not a natural culture. It's very hard to get a macho man to behave gentlemanly, right? So we're going to be more gentlemen now. Cautious, very focused on cost. So these are the five areas what we call Operational Excellence, right? Sounds obvious, right? Almost everyone will say in their mind, Operational Excellence is reduced cost. The point is it's not the only element, yeah? Okay. So we told you that we're going to save MYR 5 billion back in 2017. We have saved MYR 3.6 billion since as of today.
We are very much on track. Maybe we can overachieve that. But it's tougher and tougher and tougher, to be frank with you. All the low-hanging fruits we have taken care of. There are a lot of things we have to do, and we have achieved that. And by the way, deferment is not savings, right? So we have measured this quite scientifically too. We even have audited these numbers, right? W e told you that we're going to increase at least 300 basis points EBITDA margin by 2022. I think it's going to be easy. S o far, we almost achieved that in 2019. Maybe the target was too easy in the first place from a low margin. In 2018, we said that we want to reduce capital and CapEx intensity in three to four years. Not there yet, but we are approaching that. Now, my only caveat, - 5G.
That's a headache of it. We'll talk about it. Okay. W e want to tell you more that we aspire to do, right? Again, from an Operational Excellence perspective, Axiata ROIC must be more than WACC by 2022. That's our aspiration. We think we can do it? We're going to tell you that we're going to bump up our capability on digital or digitization analytics, not just being on par, but the best in every market that we operate in. B y now, we have really, really massively jump-started this whole process, as I said earlier. Customer promise, you will see more of smart networks. We'll talk about it later on. Celcom network costs as 2-3 percentage points improvement by 2022. Frankly, it's easy, but I don't want to, well, internally, we think we can do better than that, but externally, that's our aspiration. Same thing with margin.
Yeah, the word MFRS is the most complicated thing that I've ever had and I've ever experienced. You will see our charts, pre-post, pre-post until many of us are confused. I tell you, the numbers, we always ask ourselves, how come suddenly it's so good? Actually, it's bad because the post-MFRS margin suddenly becomes better. So you will see that. I hope you will be patient with us. We are very careful to be pre and post because it's like comparing apple and orange sometimes, right? F inally, Ibu Dian committed to me by 2022, ROIC will be higher than WACC. According to her, it's easy. Okay. So this is our aspiration. I wouldn't use the word commitment, but you know what I mean. We are going to strive to achieve those. T hese are the topics we'll be presenting today. Again, centering very much on Operational Excellence.
Agenda 8, we'll talk about how we can keep our promise while reducing our costs. Agenda 5 and 6, we'll talk about delivering the lower cost per GB. We want to be the word we use internally, low-cost producer, and yet meeting our customer promise. Agenda 6, we'll talk about Operational Excellence in general and more cost, cost, cost. Finally, agenda 9, we'll talk about digitization analytics. It's a very focused investors' analyst day this time around because we want to send a message right across. We mean business. Thank you.
Okay. Very good afternoon to all of you. I think it's an exciting time. I think what we are going to talk about is something which has been part of our DNA now.
There's a lot we've covered so far on this, but this is a journey which we'll never get over given the fact that the realization in our markets keep coming down. So the only way to survive is to keep delivering on lowest cost producer ambition, which we have. What I'll do is I'll just do a quick presentation, and then I'll call three of the OPCO CEOs/CFO to talk about what they are doing in their respective markets. I think it's very important. I thought we thought that they should talk about it because we can keep giving you all kind of things we want to achieve, but what's happening actually on the ground to deliver the ambition of Operational Excellence, it's good to hear from them.
Then I'll kind of moderate with a couple of questions and then open it up for any questions that any one of you have here. So let me start. We set up this whole MYR 5 billion ambition back in 2017. And when we said, "Look, what do we know how that's going to be achieved?" So I would say that 2019 so far, from a target standpoint, has been fairly good. We've got around MYR 800 million in the bag. We are targeting to do a MYR 1 billion saving in 2019. Now, this is obviously at a gross level because the net impact on P&L is reflected in the way overall cost development, how it happens. A large part of that saving is coming from some of our key markets. A ll of them have been actually contributing for our saving.
I must say, most of it is actually coming from Southeast Asia this year, unlike the last couple of years where we did get savings coming from our South Asian operations. Good news is, until the last 2017 and 2018, it was a lot coming from CapEx saving, which we were able to do through better negotiations and optimization there. But now we are seeing a lot of benefits coming from the OpEx side, which is actually reflective on our EBITDA margin improvement by 2.3 percentage points. And most of it is actual real saving. These are not things like cost avoidance, etc. So that, I think, is a positive direction. What we run is we run a project called ARISE, which is basically focused around delivering this MYR 5 billion saving over the five-year time frame.
I think, despite the fact that we've seen cost actually growing, specifically coming out of the network growth, we've been able to manage keeping absolute costs pretty much flat by bringing in saving under different areas. I mean, just to give you an example, we've added around 23,000 nodes in 2019 so far, which is the big red bar you see on the chart where the cost increase comes essentially out of the network expansion happening in our key markets, Bangladesh and XL, and also in some of the other markets. But that's been offset by significant cost programs being run in areas of network and the sales and marketing and some of the people-related costs.
Overall, if you look at the bar, the two blue, it's around 1%, slightly over 1% growth in absolute cost for us in 2019 so far, versus a 5% growth in the revenue, which we've delivered in the first three quarters. Let me go to the next one. I think this is the journey we had so far on delivering the MYR 5 billion savings. W e've actually delivered around close to MYR 3.6 billion so far. We've got a journey to deliver another MYR 1.4 billion. But that doesn't mean that's the end of it. I think we are now looking at what is it that we need to do more from a structural standpoint and looking at how we run the operations where we leverage our scale advantage going forward. So I think that MYR 5 billion pretty much on track for us, what we've decided to deliver.
I think what this slide actually talks about is an important element. This is where the whole cost per GB becomes an important aspect. If you see this line, the one in the red is actually the realization, and the one in green is cost per GB for us. If you look at the blue bars, they are basically saying that we've been actually growing our 30%-40% year-on-year data traffic. That does mean, and the fact that the outputs have been pretty stable, the realization for our group has been coming down. Our realization on average is now around $0.0046 per GB, whereas we've been able to get our cost down to around $0.0031. We are now getting margin, which around a year back, we were pretty much not making any margins on our data.
Most of our margins were coming from our legacy voice revenue, which was still there, whereas the investments were very little going into the voice. Now, I think we are seeing this gap widening where we are able to get margins coming in from our cost programs, which we've been running. Let me go to the next one. What does Operational Excellence from a sustainable cost position standpoint mean for us? So the first is what we are saying, sweating existing assets. Second one is about leveraging collective buying power. I think I'll cover a little bit in detail what does this mean, what we've been doing. And third one is about delivering fast, flat cost growth, and then simplification, digitization. And the last one is about capital allocation. How do we decide where to put money for us based on what the short-term gains we want to achieve?
This one, I think, sweating assets. Tan Sri did mention about the fact that how do we look at new investments, but more important is how does each of the operations look at investing into the new network expansion. So it's driven. I think each OPCO has got a different methodology, but it's driven based on analytics, based on data available to decide, not just the internal data, but also external data on where to invest to get the fastest payback. T his is driven based on profitability and payback of each and every new tower or base station which has been put in place. Second aspect is around the discussions which we have. What does this mean? Just in time and just nice.
I will not spend too much time on explaining what it is because I know one of the speakers later on will talk a little bit more around what does this mean in the context, just nice and just in time. But it's fair to say just nice is about not having an ambition of giving the fastest speed, which consumers, other than a brand positioning, have got no effect on the consumer experience. T his is about what is right kind of a speed to be given for a consumer to have the best video experience. Collective buying, I think we've been moving now to put the demand together for the entire group, looking at standardization of our configurations, looking at reducing number of configurations so that we are actually able to leverage much better negotiation powers with our vendors.
We did this exercise for the first time in 2019, where we actually consolidated the entire demand for all the groups. I mean, we buy MYR 1.6 billion worth of CapEx every year, so it's a substantial scale as a group we have. Now we are trying to build that advantage when we negotiate with our vendors, but that cannot be done just by pure negotiation. It can be done. It has to be done through standardization, reducing the number of configurations, talking to them in the same language in terms of commercial terms, leveraging a common price book, which is the same across all OpCos, with an advantage you could actually exercise in some of the markets where the vendor is more strategically inclined to further give commercial benefits.
I think that exercise we did this year, there are a lot of learnings for us, but we did manage to get a 20% overall saving for the group on CapEx. T his is based on the total cost of ownership, which is over a period of three years. So we should see these benefits flowing to us on the CapEx. And then looking at how do we aggregate to reach at least 50% of our total demand together from a common buying perspective? What does this mean for us from a cost excellence program? And if you look at what does this mean for us? I think clearly it demonstrates that our costs have just grown by 1%.
Across OPCO, you would see that we are actually getting much, much better effect on cost. Absolute cost in Celcom and XL has actually seen the growth and the marginal improvement, increase in cost in some of the other markets versus a much better growth in the revenue in those markets. These are two examples which I just wanted to talk about, which is XL and Robi. And why I talk about these two examples? Because these are two markets where we continue to invest. If you look at these two markets, both of them have added significant number of towers through whether it is a [Uncertain] strategy, which Robi had, or whether it is an Ex-Java strategy, which XL had. W e've seen that even though the growth is happening, we are able to bring down the cost per tower.
Effectively, overall cost, which I showed you earlier, we've been kind of flat when it comes to the network cost. I think this on digitization, I think there's been a journey which we've taken up in terms of moving to digital recharges. 15% penetration in 2019 from 12% now. We are targeting to move to around 30% of all charges to be digital. We are looking at significant bump in the number of customers acquired through digital routes. Digital care, I think we have nearly 15% of our I mean, at this point in time, around 25% of our calls in our call center actually ends up with the agent. Our target is to bring down to that to 15% of those calls and also double up the mobile application usage. Automation, I think we started very early with very few bots at the beginning of this year.
Now we are around 25 bots across the different operations. We expect these to be automated to the extent of 250 bots by the end of 2022, which means a lot of these would be done through an automated means and not involving significant headcounts to be involved. We are also exploring now how do we move to shared service across our operations and leveraging the scale advantage for our benefit? Similarly, on the area of network investments, I think we talked about just nice. We also look at new growth areas, which is on the investments in enterprise digital VAS. CapEx allocation, I think we've been extremely. I think Tan Sri did talk about it. How do we look at new investments? I think these are a lot driven now based on philosophy of payback and immediate impact on the profitability.
Even though on a long-term IRR, these may be attractive, but we look at it on a very short-term basis at this point in time, driven from a profit and a cash focus. We also keep aside around 15% of CapEx that is not allocated to the OpCos. These are allocated based on performance and the need of the market. Then we decide how to allocate. I mean, if you look at this year, what we guided as our CapEx for the full year, we are expected to be below that. That's essentially coming out of the fact that we do not allocate the entire CapEx to the OpCos and is driven based on the results and outcome we want to achieve. There are obviously other opportunities, which has been expressed here, we are looking at from a capital allocations perspective.
I think this is what we intend delivering on our end, and I think Tan Sri did talk about from what we committed earlier in the year in terms of the cost savings. I think MYR 3.6 billion, fairly confident by the end of next year, we should have achieved MYR 5 billion, which does mean what does it take us to go to the next stage from a cost optimization? EBITDA improvement, I think we've delivered around 2.9% so far after we said. I think aspiration is to move up, at least very clearly define agenda in Celcom. The last one is we look at our CapEx intensity. We expect that, as Tan Sri said, over time, we should be down below 20% despite the fact we continue to invest in XL and Robi. That's it from my side.
What I will do is now I will invite three of the gentlemen from three different OpCos.
Good afternoon, everyone. Let me just touch a bit about what we are practicing now. Some of these are some of the new practices in Celcom. We're going to talk about network excellence. Before that, let me just give an overview of the network. I think most of us probably know, whoever that's using Celcom today and those that are not using Celcom probably not experienced this, but those that are using Celcom know that Celcom now is delivering the widest coverage, one of the most consistent and reliable networks in the country. Overall, we have about 98% population coverage, and when we talk about LTE, we have about 93.2% population coverage in terms of it.
We have built this network, and we have built this network more to give the best video experience for our users. I think some of this, if you look at some of the other stats, we still have 2G network, 3G network, LTE, and also LTE+ . Our 2G network is always about 95%, 94% of 3G, LTE 93%. And LTE+ , actually, is already reaching about 81.3% in terms of population coverage. If you look at the three bars on the right, it's something that we use more frequently, is the report by Opensignal that looks at some of the performance of the network. These are the three areas. There's a few other areas that they look at. If you look at from this, the Celcom's network is the highest in terms of 4G availability in the country. It hits about 87.7%.
Celcom's network today is also in terms of the best in terms of video experience. Here is about 63.5% video experience. Also in terms of latency, Celcom network is also the best in terms of latency. This is the lower the better. They also look at in terms of download speed and also upload speed, which is Celcom's network. It's not that it's not important. It's still important that we fall actually one of the top two in the market. Overall, in terms of performance, we have built a network with the widest coverage, consistent performance, and built for video experience. T his is, in a way, verified by these third-party sources, which is based on crowdsourcing. While we're building all this, I think another thing that probably I would like to share with everyone is what's happening in the network itself.
Number one, everybody knows that today's network is no longer about a voice network, but I'm going to tell you, today's network is also not just about data network. Today's network, actually, we are running a video network. If you look at the pie chart, that's the overall total of data traffic that we are running through our network, and then 60% of this traffic is actually video traffic. So 60% of the total traffic that's running data traffic through our network is actually video traffic, and 75% of the 60% is actually YouTube and Facebook. So actually, YouTube is one of the largest in terms of more traffic that's going. So this is why it is very important for us now when we look at designing the network, operating the network, dimensioning the network. It's not about just dimensioning a data network. It's actually dimensioning for a video network.
The video experience is very important. If you look at some of the stats in terms of the traffic growth, the overall data volume has increased more than 20% in the last nine months. If you look at the video traffic growth, also more than 15%. And data usage per sub also has increased more than 30%. Now we're talking about 14-15 GBs per user per month in terms of the utilization over the network. So what do we do when we run a video network? T here's a few things. There are a lot of things, actually, that we do. These are just some of the top three things that we actually do in terms of how do you optimize the network. Number one is some of the tools that we put in place.
Of course, they're about digitalization, automation, and analytics tool that we put in place. Of course, we embrace the digitalization in network planning, performance management, service quality management. Those tools are put in place. We put more advanced deep packet inspection tools and also predictive and prescriptive analytics. This is how we're starting to put artificial intelligence into our network planning. And then, of course, multi-vendor performance, etc. But one of the bigger things that we do is also we're starting to look at into differentiated quality of service is depending on either service-based and also user-based. Customers can decide to use by the network differently and use the network differently. So we can optimize this based on that differentiation. And then also further optimize the video traffic using the video optimizer, etc.
So differentiated QoS is one thing that actually will help in terms of how we manage the traffic and the network better. T he third item is about crowdsourcing. We are not looking just at network in general, but use a lot of crowdsourcing data. This includes, for example, using Ookla. Ookla pretty much is download and upload speed. But we also use Opensignal that looks at beyond that. Opensignal look at download, upload, latency, video experience, 4G availability, voice quality. Then we also benchmark this with our competition, which we work with our competition. Then we look also this cluster by cluster. It's not just overall network. We look at cluster by cluster. This is an example. For example, we did an example for Sabah. We look at the performance each cluster by cluster, and then we rank them.
If we were number one operator in that cluster, we dimension it differently. If you were number two operator in that cluster, then we upgrade differently. So we look at this from cluster by cluster, and Tan Sri was saying earlier, it's down to also side by side where we can. So it's just, look at number one is crowdsourcing data. Number two, look at cluster by cluster. Number three, we also look at what the competition is doing in the area. I t's not just standing on our own. T hese are some of the three things that we do, how to optimize our network utilization from digitization in terms of tools, looking at differentiated QoS, then crowdsourcing and ranking and prioritizing the upgrades, etc. With this, we hope to see, for example, the overall network operating costs will decline.
If you look at from cost perspective, the network OpEx may be between 10% to 15% within the next three years. Significantly also on the cost per site, more than 5% into our savings on cost per site. More importantly, I think the cost per GB, more than half in the next three years. Of course, there are other things, as I mentioned, that we do, how we look at some of the low traffic sites, how we're insourcing some of these activities, even energy and power management, we're looking at site to site. Tan Sri was mentioning about using gensets and batteries and double and triple redundancies, etc. A lot of this that we hope is one of the in terms of operating costs will reduce over the next three years.
Just to give you an idea, I'm not going to go through one by one the detail of this. Just to give you an idea in terms of when we prioritize for capital expenditure for CapEx. There's a few things that we look at. For example, the first is in terms of the investment triggers and also the way we dimension the network. Tan Sri Jamal mentioned earlier, for example, in between 360p and 720p and 1080p, if you are on a small screen, you probably won't be able to see the difference. But if you look at it, if you look at it from a video quality perspective, if you want to give a 480p in terms of quality, you only need actually a three megabits per second speed to give to the device. And also depending on the device and the device quality of the device.
If you buy the latest device with the highest resolution screen versus the device that actually has a low resolution screen, you probably won't be able to tell the difference. I f you go to 720p, you only need actually 5 megabits per second. And if you want to go 1080 and 4K, you only need 8 and 16 megabits per second. So we're starting to measure this. And then the way that we trigger the upgrade of the investment of the sites is depending on whether we are delivering this kind of quality to our subscribers and depending on what they pay for and what the services they use. That's kind of the first step that we look at. The second step is when we want to upgrade, for example, there are a lot of steps that we do before we add the new sites.
There's, for example, adding up. If you see there, quite small, but it's L18 base layer. We add L18 upgrade to more spectrum. And then if needed more, we upgrade to L2100. If you need more, then you add on 2600 layer, etc., and so on and so forth. After all that, then only you add new sites to add on to the new sites. Even that, we also look at the next stage, which is look at, hey, even if we add a new site, what's the site profitability in that area? Are you looking at whether in the area is a margin enhancer or is it a profit contributor, etc., etc. So we have the kind of criteria that we look at before we decide even to put a new site into that location.
Last but not least is the process internally that we go through is the Network Investment Committee that will decide where to prioritize some of these investments and so on and so forth. T his is just an example of some of the steps that we put in place in terms of how to optimize every CapEx, every new site that we're going to put in to enhance the services that we give to our customers. Now, as a result, this is our CapEx. So, for example, we look at 2020, for example, based on delivering of this quality of service, it's the first time that we're looking at actually we are not spending as much on Radio Access Network as we used to. Because there are many things that we do before.
If you look at the inner circle there, it's for the first time, if you look at it, that the actual RAN is not the highest part of the network CapEx investment. So we're looking at about 40% savings in terms of RAN CapEx. So what we can do, what it means is we can spend a little bit more on strengthening of the network. If you see the green part of the bar is how we strengthen the transport, how we bring more fiber to the sites, how to increase the number of fiberized sites or single hop to fiber sites. T hat's how we can allocate our capital investment better to deliver the service.
By changing the network triggers, if you look at it using the existing triggers versus the new triggers that I mentioned just now, we're looking at savings of 40% of CapEx savings just from relooking at those trigger criteria by itself. T hese are some of the things that we do in order to optimize the investment, optimize the CapEx investment. Vivek, so this is the end of it. But just to summarize, we have built the widest and the strongest network in the country. W e've been doing this for more than 30 years, actually. W e've been here for more than 30 years. The network is now almost ready for 5G. The network now has evolved from a voice to data to a video network. W e're actually running a video network. W e're optimizing our network for video experience.
There are a lot of tools that we've put in place. We use the term just-in-time as just-in-time network investment. L ast but not least, I think what I mentioned just now in terms of network OpEx, we are looking at either 2-3 percentage points reduction over the next three years. A s a result, also in terms of EBITDA margin to improve by 3-4 percentage points by 2022, which is if post-MFRS 16, it will be in the north of 40% EBITDA margin. Thank you.
Yeah, thank you, Idham. I think we'll go to the next speaker. That's Adlan.
Afternoon, everyone. I think in line with our topic today, I think we are going to probably share how in XL, how do we look at our network rollout strategy, focusing more on profits and returns.
I think if you see the target that has been set by our shareholder, it's for our ROI to be more than WACC by 2022. So far, I think what we've seen is we are probably on track to achieving that. Just a snapshot of where we are in 2019. 2019 has been an exceptional year for XL. So if you look at double-digit revenue growth, EBITDA growing by 19% in line with Operational Excellence, EBITDA is growing faster than revenue. I think you see at margin growth 36%-39%. And I think all this is in line with our aspiration to drive for the lowest cost per GB. If you look at this year alone, I think we have managed to lower down our cost per GB by approximately 30%.
We are also contributing quite significantly to the overall Group Cost Excellence Project, where I think up to September alone, we have actually delivered close to about IDR 1.3 trillion in terms of cost saving. As a result of all this, I think you will probably see that our returns. We are back to profit, and I think our free cash flow has also increased quite substantially. The key question is how do we sustain all this and how do we even improve this moving to next year and the year after?
The reason why we are able to achieve this is, I think, for us, we have been following a strict end-to-end process framework to maximize our return and profit from the time when we actually roll out our network to the point of time when we optimize and expand our network for capacity to the holistic GTM process where we get all the units, be it finance, be it network, marketing, sales, dealers with a single KPI focus. That's something that we have really done well, I think I would say over the last 12-18 months. Last but not least is a very granular monitoring and tracking of our cluster and power performance, where I think this is done even on a daily basis where I think that's visible to all the business users and the salespeople on the ground.
We'll take you a little bit on some of these key elements and talk you through this in subsequent slides. This is very important. I think we probably want to share a little bit about our strategy in terms of our network rollout philosophy. While I think in line with what Tan Sri said, we don't want to build the best network. But in building our network, I think we focus in trying to see what would give us the maximum ROI. H ence, when we build our network, we actually build an average network, a network that is actually good enough to be able to compete with the market leader. I n our investment philosophy, you look at when we attack into new areas, which is our source of growth, revenue growth in 2019. T hat will continue to be our source of revenue growth moving forward as well.
We will continue to explore in investing in new coverage area, which I think we see vast opportunity. But at the same time, we also see that there's a huge investment required to expand in some of these areas. And to a certain extent, some parts, some clusters may require longer payback than we would probably have anticipated. Anyway, I think when we do all this, we would actually balance into our investment philosophy in new areas between returns as well as payback. At the same time, I think we will probably balance in terms of how do we invest in terms of capacity. W e actually, in line with the Operational Excellence, we have actually put in a more granular and probably tighter measures in terms of rolling out for capacity expansion in some of these clusters.
It's not necessary that once it hits a certain trigger that you probably put more capacity. But I think typically we will look at other measures. We'll go back, we look at the cluster profitability, we look at the yield, we look at the subsequent event that's probably involved. F or example, if that cluster is not giving us a good return below a certain yield, for example, the marketing has got to really go back and probably see whether we could actually monetize before we actually decide that we'll probably put in more capacity or CapEx in that. What is actually more important in all this, whatever we do, we will probably benchmark against our network with the competitors. In that sense, we need to make sure that we achieve a certain level of quality that we are able to compete with market leaders.
We will not be able to match them dollar for dollar. So our network in some and most of these areas will be slightly less superior than the market leader, but so long as I think it gives us a good alternative to compete with the market leader, we have seen that this has been successful to provide second-best alternative but at better value to consumer. W hen we build our network, it's very, very clear, we are actually building a network more on reliability rather than speed. So if you look at the right-hand side, I mean, typically, as you see, we look at IRR, we look at standalone things with the highest return, but we've actually introduced payback as one of the key criteria and measures today in terms of deciding where we actually invest and where we prioritize our investment.
Not forgoing the long-term goal, but also with a clear short-term focus as well in mind to deliver some of the short-term metrics. All this, I think, when we come up with the network rollout plan, I mean, we look at it quite granular. We break down the Indonesian cities to about 514 cities to various kabupaten, clusters, as well as regions. I think when we do our network rollout, it's all backed by data analytics. We use a lot of data analytics, be it internal data, be it external data, whether it's in terms of GDP, market value, ARPU penetration, population, and opportunities on government expansion plan, for example, are some of the key measures that we put as part of our evaluation of our investment.
Not only that, but we overlap even the Facebook data, the Android in terms of data traffic, as well as YouTube data speed, or even the network quality from Opensignal. Based on all this data, I think we come up with an investment matrix, investment matrix, where I think we divide into key quadrants. By the way, these numbers that I've put in there are all dummy numbers. It's not really the numbers. But you can use that as a rough guidance of what the real number is. But anyway, so once we have actually grouped the clusters, the cities into these four quadrants, we will actually prioritize the investment based on revenue, investment, as well as returns and payback that we've actually introduced.
So we rank each and every one of these to decide where do we draw the line and which area do we invest to achieve all the metrics that I think we have defined for. So one of the things that we have noticed, moving from building an average network, moving, let's say, from a below average to an average actually gives you the fastest return. Let me give you an indication of what is an average network. We are looking at around 1 Mbps average speed. We are looking at probably a population coverage, 4G population coverage of not less than 90%, and a gap of not more than 5% against a market leader.
Typically, based on this investment philosophy, we see that moving a cluster from below average to average actually gives us a faster return, be it from a revenue standpoint or be it from a return standpoint. Hence, I think when we focus in terms of deciding in terms of our where to invest, we'll be focusing in bringing our below average cluster to average. Today, I would say that more than 60% of our clusters are categorized as average or above average, average and above. By the way, this classification is actually quite dynamic because it's pegged to where competitors are. If competitors are moving, we can be, let's say, average today. If competitors are actually spending more money and as a result have a better network, we may drop to below average.
This is a consistent measure that we probably have in terms of looking at how we look at returns in the various clusters that we have. This is one of the key guidance that we will, principle that we will probably use in terms of investing or rolling out to new cities. We actually have a very comprehensive process, which we have set. When we decide where to invest, we have set a target for our revenue to achieve certain return profiles as well as the clusters. Once we've done that together with marketing, finance, network, as well as the sales region, including dealers, we will come up with a holistic plan, for example, to attack the market. We know, for example, there is typically a time lapse between when the network is ready and when I think when we need to create.
So we typically create three months before the network is ready. We already create awareness in certain clusters or cities, for example, to say that we are coming to those cities. And I think with all this 360, everybody across divisions has got a single focus KPI to drive tower profitability and cluster profitability. We actually have set a target what are the revenue targets or clusters or revenue that we see per tower basis for each and every one of the clusters that we distribute to all the marketing, including the regions, as well as the salespeople as well and dealers as well. And all their commissions are actually tied to this. I t's not just holistic go-to-market, but also the KPI and the reward are actually linked to this tower and cluster profitability.
A t the end of it, we will actually have a regular monitoring of this tower profitability and tracking. I n that respect, if tower doesn't achieve a certain level of profitability by certain times, we will actually relocate the towers. I think this is also in line with our cost initiative as well. When we negotiate with the tower providers, for example, today, we actually ask for 12-15 months of rental holiday. I n that respect, if within 12 months we don't get certain tower revenue or we are not even able to break even to cover even our variable costs, we will actually relocate the tower. A ll these things are quite, I would say, aligned and a comprehensive process that we have. Last but not least, so far it's been delivering good results.
I think if you look at our tower profitability, be it Java, be it non-Java, it's actually growing. And if you look at on average, I think our Ex-Java revenue per towers are actually growing single digit. But I think the Ex-Java revenue per towers are growing quite substantial, I would say, in the numbers. A s a result of all this as well, with the better methodology, with using a lot more data and analytics, we also see that we actually are able to shorten our payback period. So this new investment that I have used, this framework and this granular process and close coordination across division, we've seen that our payback actually has come down to about three-to-four years. So so far it's been going good results.
Of course, with the Operational Excellence, we probably need to tighten and actually improve on the framework that we already have and improve the coordination that we have across organization.
Thank you, Adlan. Now I'll ask Thomas to talk about what's happening in Smart. Smart is one of the companies which has had Operational Excellence for quite a while. So yeah, Thomas.
Right. Good afternoon, everyone. It's a bit of a rare occasion that you can get some extra insights on Smart. I think I presented two years ago. Happy to share some of the success stories at Smart today. T opic of my presentation, my discussion is sweating assets to drive cost per 1GB down. The Cambodian market historically is characterized as being one of the lowest priced markets you can find in the world, probably. Headline pricing is down to $1 gives you 8GB of data.
Even though it has eased out over the last, let's say, 18 months a bit, the market got a bit tired from price wars. It's still one of the rock-bottom markets, and that obviously requires us to be extremely sharp on how we stay profitable and how we grow our revenues. We have seen, as a result of that price war, a tremendous amount of uptake of data traffic, in particular 2017 to 2018, a huge spike, so the market is elastic. Consumers, they are craving for data because certainly in Cambodia, we don't have such a well-built-out fixed-line infrastructure. So the majority of data consumption goes mobile. We have now actually hit almost 15 GBs of data per data user on average in the months now, in October and November. Overall data traffic has gone through the roof.
At the same time, our revenue mix has fundamentally changed over the years, so we are now at year-to-date 64% data revenues, almost hitting the 70%. That obviously gives us a big challenge, namely driving the cost down, and happy to present that Smart is probably the OpCo, is the lowest cost per 1GB in the group, and yeah, we had $0.13 a GB year-to-date. In Q3, we were down to $0.11. So we are almost at the $0.10 threshold, which I believe is a going-forward benchmark we in the group have to hit for, so as you can see, despite that very low yield, below $0.30 GBs, the margins are actually healthy, and the margins, in a way, are actually even widening despite a further decline of yield, even though the yield decline has flattened a bit. So what is the secret sauce?
I mean, in Smart, cost excellence and cost consciousness has been quite embedded in our DNA from day one, but certainly, we have a few aspects which enable us to be such a low-cost factory, and one of the reasons is certainly we are not driving for the fastest network. Yes, there was a certain moment of time when Smart launched 4G ahead of the competition that we were proudly crowning ourselves as that speed crown, but since then, actually, it was never the purpose to be the number one in Ookla or the number one in Speedtest. Actually, it is all about offering a consistent experience to the users, which is good enough for exactly their use case, and yes, certainly, they're watching YouTube and Facebook predominantly.
We need to make sure that we are delivering a good video experience, but it not has to be in 4K or in even HD resolution. As you can see, the network speeds are hovering in Smart's case somewhere at 15Mbps per second, actually. We are not aiming now to shoot up and going far beyond. We are relatively blessed in Cambodia with a good chunk of spectrum in Smart. We are owning roughly 100 megahertz of prime spectrum. That allows us certainly to enable a lot of carriers for each site. Smart is operating at the moment 2,700 sites. All of them are equipped with 4G. Yes, 3G as well, but that's more of a side product.
We are looking intensively, obviously, at how to make sure that the maximum of spectrum we have goes to 4G so that we can run at a maximum of capacity per each site. That brings down the costs. We are managing our network in-house. For us, outsourcing just for the sake of outsourcing was never a question. We have not seen an opportunity yet to yield better costs in a way if we operate our network out of house. We are using local subcontractors instead of going for the full turnkey approach, for example, when we roll out. We are running, yes, on the latest of equipment. We are using 8T8R Massive MIMO.
I mean, in a market like Cambodia, where you would eventually believe, "Why do you need to go latest tech?" It makes a hell lot of sense for us to go latest tech because the efficiency is coming out of it. We are constantly reshuffling the network to make sure that wherever we need high capacity, we have high capacity. Where we don't need it, we are going lower capacity. We are very active in migrating consumers to 4G, Voice over LTE, Voice over Wi-Fi to enable as much as spectrum for 4G data at the end of the day. Core network is top-notch, latest stuff, ready for 5G eventually. And yeah, so there are a lot of measures in the game where there's probably not one single measure which is the killer measure besides having so much spectrum that we can really run high-capacity sites.
There is a bunch of things on the network side and the tech side which make us shooting capacity-wise per site to the level we need. You can roughly see that traffic per site here in that slide is going very well up, up, up. That brings the margin in which we need. So we have roughly 900 GBs of traffic per site day. We have some sites which are carrying 2,300GBs or 2,400 GBs per site on a single day. Putting this into comparison with other markets, you can see that Smart is serving roughly 20,000 GBs per site in a month. There is no any other market which we are aware of, Ncell in Asia, not in Europe, which is delivering such a high traffic per site.
In a way, cost factor is a result of traffic versus, yeah, I mean, the load you carry on the site. Then the costs actually are relatively static. The more traffic we channel through, the lower cost we get. We will not rest at where we are. We have a lot of measures still in play. As I said, it's not the one secret source saying, "Okay, we will bring it down further." There are a lot of individual measures we are working on, such as interference management, such as power management using self-optimized network stuff, using video optimization, TCP optimization equipment and solutions, RAN signaling, buffering, etc. We believe that we can bring the cost per gig down a bit further. We will certainly hit a threshold.
There is a base cost which we can't overcome, at least not in 4G, which probably sits in Smart's case somewhere at $0.08-$0.09 per 1GB for now, but we are aiming to get there in 2020. Looking ahead, well, as a result of still optimizing further and being so cost-efficient, despite such a very low price line, we are running at roughly 12% network OpEx and yielding on about 50% EBITDA and 25% PAT, despite regulators also in Cambodia asking constantly for additional income from their end, well, I mean, now 70% of our revenue coming from data. We see voice is still tapering out, and based on our profitability on data, we believe that in the next years, we can actually edge up EBITDA a few couple of percentage points more. Thank you.
Thank you, Thomas. Thank you, Adlan, and thank you, Idham.
I had a few questions which I wanted to put these guys on the tight spot, but I realized there are enough questions from the audience which have already come in. So let me start with these questions, and the first one is from Ranjan, which says, "How much of strategy of reducing unit data cost is predicated on spectrum? Where's it gone? Spectrum and network sharing. Can XL Axiata pursue network JV in any of its markets?" I think the first one, I think what we've done here is mostly business as usual, so it's not factoring in any potential opportunities of new spectrum acquisition or network sharing, so that potentially would be an upside if we are able to do that, and the reason why we have not factored in is, first of all, spectrum acquisition will always be determined based on the business case.
Is there a value of the spectrum supports that? N etwork sharing has been always a question mark given that fact that there are regulatory constraints in some of the markets and also the fact that you have to have someone on the other side willing to do that. So again, there are obviously opportunities of network JV in, for example, Malaysia could be. But some of the other markets, regulatory constraints actually prevent us from doing that. And we've seen that, for example, in Indonesia, Bangladesh, etc. I think that's on the first question. Second one is from Piyush. Adlan, maybe you need to answer that. "In XL, XL Axiata changed in strategy to focus on shorter payback period for network expansion. Is it driven by your learning in Ex-Java network expansion or in sharpening focus based on those learnings? Can you elaborate on any such learnings?"
Hello. Okay.
So I think going back to the question, right? So I don't think we are changing any of our strategy, right? Investment outside Java is still something that we are going to pursue. I f we look at the results that we have seen so far, I think these are our source of growth. But what we are actually doing in line with the Group's Operational Excellence, we are tweaking a little bit in terms of how we execute this, right? I n the past, we just purely looked from a return perspective. But here, we have overlaid the return on paybacks, right? W e've set ourselves a threshold of what sort of period that we are willing to take, right? I n the past, it was probably longer, between 5 to 5.5 years.
But given what we've seen outside Java today, right, 3 to 4 years, we can get quite decent. We can get that sort of payback. I n all our new investment outside Java, we have set a shorter time period horizon for us to realize the payback. O bviously, that means that not many new areas that we are able to hit. But I think the focus, given the short term and the balancing act between long term, is something that we are probably going to pursue, especially in 2020, not forgetting the aspiration and the strategy of investing outside Java.
Okay. Thank you, Adlan. What's the next question?
It's taking time.
So then I'll ask Adlan a question. Adlan, ROIC greater than WACC by 2022. Once you put that on a slide, how do you think you will make it happen?
Yeah, I think, again, right, I think going back to the framework that we have, right, we are not going to build. We are going to build a network of an average network, right? We are not going to overspec our network. W e are going to build a network that's good enough to compete with our competitors, right? I n this case, it's going to be Telkomsel in most of the clusters. H ence, I think because of the more ROI focus in terms of building this, I think we are able to realize our investment payback probably within a shorter time period as we've seen, for example, in the last three years in terms of the numbers, right?
In addition, I think one thing that we have done extremely well as well is the cross-functional working relation across marketing, network, finance, the regions, the salespeople, and the dealers as well. T hat will actually give a single source of KPI for everyone to focus on, right? I think with this as well, and we are able to drive the troops on the ground once you actually have a network ready and the towers are actually ready as well to actually push for sales, driving for more retail outlets around the towers and all that, right? B ecause of the fact that the performance of the towers today are also visible to all the operation people, including sales and dealers, that will actually, and that's all linked to their KPI and reward, that also drives their motivation to go out and actually load up all these towers.
We have started doing this for the last 12 and 18 months on this, a nd we see that it's actually driving really positive results. H ence, based on that, I think we are quite optimistic given the tight framework that we have. Using more data analytics, we are able to spot on the right area to invest. We are able to hopefully achieve the right versus WACC by 2022.
Okay. Let me then come back to the questions already there. Maybe Idham, the first one from Piyush. What cost efficiency as a percent of revenue and CapEx efficiency can be achieved with potential network sharing? T he question is, is there a potential opportunity for network sharing in Celcom? And also, what's the optimum 5G network sharing model which Axiata would like to embrace?
Okay. As you know, we are running the 2G, 3G, and 4G networks. We have looked at the cost of running this network. If you look at the 2G networks, actually, when you look at it deeper, the cost of running 2G is quite minimal compared to the kind of revenue you are generating on the 2G network from the voice network that is coming into the 2G network. So we have looked at that. It's actually better for us to just maintain that network and we have the widest 2G network in the country, so it's more that can we be able to carry more traffic into the existing network. I think it's more of that from that point of view. But 3G network, I think that's the one that we're looking at is to answer the next question as well. Is it better for us to shut down the 2G network? Yes, this is something that we are studying.
We're looking into it because the traffic that 3G is carrying today is predominantly the voice traffic for the 4G network. T his is something that we're looking at. That leads us to the 4G network. Are we able then to share this 4G network? What makes sense to us is to look at the sharing of the 4G network, especially in the area of suburban and also the rural areas. T his is something that actually we are seriously looking at because I think we think as an industry, this is one area that especially to achieve the target of the NFCP target for the rural areas, the 4G network in the Zone 3 and Zone 4 is where we need to share. W e are planning to share. So this is cross-industry. We're looking at this.
I'm sure there is a lot of CapEx efficiency, but the mechanics of how this is to be shared is still being worked on by the industry. I don't think we can even give you a number, but I'm sure it's quite substantial in terms of how we do in these areas. Today, if you look at it, each of the operators have about 3,000-plus towers in Zone 3 and Zone 4. Yeah? So Celcom has 3,000-plus, Maxis is 3,000-plus, same as Digi. So you don't really need more than 4,000 towers overall to cover the whole Zone 3 and Zone 4. So today alone, we already have 9,000-10,000. T hat savings is quite significant in Zone 3 and Zone 4. Z one 1 and Zone 2, I think that's we agreed is to leave it more of a commercial in nature in terms of sharing.
That comes to sharing. On the 5G, I think you probably have read that we signed the MoU with Maxis last Monday. At this point, it is to explore in terms of for the city centers and urban areas how to share the network, how to build this network together as opposed to we go and build separately and then try to share later. We're trying to plan for the sharing in the outset, in the onset, right? Earlier in the process. To what extent and how and which model are we going to use? This is something that both parties still need to work on, right? Because given that the 5G technology is still, the standard is still not finalized when it comes to sharing. One of the reasons why we decided with Maxis is because we are using a common vendor.
So it's a lot easier for us to plan for the sharing. So this should be finalized within the next six months or six to eight months. But we believe that the savings will be substantial.
Yeah. Thank you, Idham. I think just quickly, maybe there were three questions which together, Adlan, you can answer. One is the first one, to what extent is the group willing to fund CapEx to move from below average to average network? That's one. And I think second one also is there any plan to shut 3G in Indonesia? And then the last one, what are the costs associated with relocation of towers for Indonesia? Are they significant? And how fast is the relocation process post-termination of the existing towers?
Okay. I think if we look at moving from below average to average, I think that's going to be the crux of our investment, right? Because that would probably give us the fastest return, right? So fastest and probably a shortest return as well. S hortest payback, sorry. A s a result of that, I think if you look at our investment philosophy, the focus is probably to bring a lot of our below average city to average. I think if you look, we still do have quite a number of cities that's falling below average. I think the focus that we have today is to focus on bringing up the cities.
At the same time, I think the balancing act in terms of allocating investment into expansion of capacity and all that, we will probably be a bit tighter in terms of expanding additional capacity in some of the areas that we are operating in, given the fact, especially if those clusters are probably not profit-making and yields are actually extremely low, right? So there is a balancing act, but I think very clear in terms of Operational Excellence, the below average moving to average will be one of our key focus. Second, shutting down 3G network. I think this is more we have a very scarce spectrum, right, to probably cater for the 60 billion or more of our customers moving forward.
Hence, I think the faster that you are able to migrate the 3G to 4G, I think you are able to free up more spectrum to cater for this additional traffic coming for 4G, right? Hence, I think, yes, I think we will accelerate that, but primarily driven by the fact of shortage of spectrum on 4G moving forward. Of course, there will be new spectrum that will be available and all that. But as a matter of fact, the cost would also be extremely expensive given some of the benchmark they will look at. But at the same time, we also know that I think assuming that if, for example, those spectrum are not available, we actually need to prolong the life of our capacity, for example, to be able to handle that increase in traffic.
I think the 3G shutdown completely will probably take some time. But at least the process to accelerate the migration from 3G and 4G is something that we'll be focusing on, right? Not only for cost saving, but also to drive the spectrum efficiency. Last but not least, on tower relocation, I think that's something that we are doing on quite a regular basis, right? However, glad to say today, based on our data analytics, typically today we get a lot of the tower that we planned for right the first time, right? But even that, I think we are not 100% accurate, right?
But given the fact that we are able to negotiate with the tower providers such commercial deal, free holiday rental for 12 or 15 months that allow us to relocate to a new site location, that actually helps in terms of us monetizing in terms of driving tower profitability as well. The cost of movement is not much. I think if you look, it's less than IDR 100 million. Typically, I think you are probably around IDR 70-80 million cost of relocation. Time-wise, if you are looking at an existing co-location, that's going to be immediate. But I think if you talk about a new location, then that's typically going to take about three to six months time. I think that's an ongoing process that we have adopted as a regular mechanics to make sure that we maximize our tower profitability. Okay.
Let me, Adlan, I think the first question was actually addressed to me. T hanks, you took it. This was to what extent is the group willing to fund CapEx to move from below average to average network? Okay. Now, I think the Indonesia XL is around one-third of our CapEx. Group CapEx is spent in XL. I think we've seen benefits of our strategy of Ex-Java. I think we are seeing returns coming from that strategy. So I think from our perspective, what is a priority at this point in time is what has been invested, let's get the best out of that investment which has been made in Ex-Java to get the best returns coming in.
That does not mean if there's a strong business case for them to invest just beyond where they are or also improving below average to average network and they have a strong business case which is driven based on certain parameters which we talked about, payback, profit, cash flow impact, etc., then we are open, we are flexible to look at that. But that's business case, capital allocation issue, whether it's this Indonesia or versus somewhere else. So it's a capital allocation, prioritization, and business case fundamentals. I think that's critical for us. If there is, we are willing to invest in. Okay. Then there is, I think there are a few more questions.
Apologies to be vague. Just want to remind the floor, there are actually two roaming mics, one on each side. If you do have any follow-up questions also, you can raise your hand. Yeah?
Cla re, I was wondering if we continue here, we will not be able to do any other sessions which have to follow. Should we park some of these questions maybe to the end as part of the Q&A instead of continuing here and then?
We can, but we are actually ahead of time.
Oh, then it's okay. Then we can pick up .
We got another 25 minutes. Plenty of time.
I thought you wouldn't say that.
Sorry, we're going to keep you up there for another 25 minutes.
Okay. Then it's fine. So I think we've covered the first one, network infrastructure spends.
Maybe the first two, yeah? What has been the spend in 2019? And in 2019, let me backtrack a bit. In 2017, 2018, and some parts of 2019 has been the expansion of the LTE network.
We've been investing a lot in terms of new areas and how to bring coverage of the LTE network. T hat's what we've been doing. A s you have seen in some of the numbers that I presented earlier, in terms of coverage now, it's population coverage. A lso in terms of availability, we have reached our target and we are now ahead of even the competition in terms of coverage. So 2019 has been onwards moving forward, as I presented, that we will be spending less on the capex towards building up new sites and expanding the availability. Not to say that we stop, right? But actually there are areas that we now move our sites, etc., we still increase our population coverage, but more towards increasing the capacity, filling up the gaps, etc.
So moving forward, it will be more of towards strengthening the network, adding up more capacity into the network. If I go to the next question from Foong, it is in the 5G deployment. What is the degree of fiberization? If you look at where we are today, there's a few ways to look at the fiberization. Number one is just a single number, how many sites are actually being fiberized, right? But if I give you a little bit more insights into it, yes, we have a number of 30-odd% in terms of our sites. It's actually today; it's already fiberized. But if you look at it from the collector side, we have more than 90% of our collector sites actually fiberized. T hen if you look at another number, how about those sites that are single hop to fiber?
Even that we're looking at now in the north of 70%, 70%-75% of our sites is actually a single hop to fiber. T he question is, given the evolution of the microwave technology, do you really need 100% of the sites to be fiberized to be ready for 5G? I mean, this is something that we are experimenting. We're looking at the E-band microwave, which can give even up to 10 gigabits per second in terms of microwave, which could be used for the access. Those are things that we are looking at. But at the same time, also we are still increasing in terms of fiberization. W e hope today that the single hop to fiber will increase from the 70-odd% to the north of 80% within the next 12-18 months.
Okay. Next question, I'll take it.
I think when you're talking about cost saving, what is the balance between cost and profitability? I think clearly profitability is the objective. It's not just cost saving. I mean, I think also as when we define Operational Excellence, it was very clear that the Operational Excellence is just not about sustainable cost position, but it's also talking about profitable growth, and it also talks about customer experience and digitization, etc. So it's to drive profitability, but it's very clear that the top line still takes a while to build, and driving top line to drive profitability is important, but it takes a while to build. Whereas cost saving is something you can actually be, and there's nothing to lose on cost saving, so there is clear focus around cost per GB, but that's in the context of realization per GB.
So we want to make margin on our data which we offer. T hat's, I think clearly profitability is the end objective. Cost is the means to drive profitability. When we can, we expect Celcom numbers to significantly improve, especially cost and bottom line.
That's what we are focusing now. I think they're focusing on the cost, especially the network cost, sales and marketing cost, and also the operations cost. But I think the guidance that we have given earlier, three-to-four percentage points in terms of EBITDA margin, we're looking at within the next three years. W e're also looking at, if you look at post-MFRS 16, it is to be in the north of 40% in terms of EBITDA margin.
So before I come to Ranjan's question on XL, ROIC, ceiling, WACC, I think I would like Thomas to at least say, talk about 5G use case.
I think it's a very unique position you are in versus where everybody else is struggling in terms of justifying a 5G use case. But you seem to be in a unique, different position. T he way you look at 5G is very different from how others have been looking at it. So I think maybe worthwhile answering this.
Indeed, we at Smart or in Cambodia in general, we believe that there is a case for 5G even relatively soon. W e are planning to be in 2020, live with 5G. The reason for it is, yeah, on the one hand side, we don't have that vast fiber infrastructure which allows consumers to hook on fiber at home or also at the office. So we believe there is a use case for 5G as a FWBB or fixed wireless access technology for homes and for offices.
At the same time, we see provided that the spectrum will be given to the operators at an affordable price point. That's obviously the caveat to it, that with 5G, we can bring down the production costs of data further, utilizing existing towers, utilizing existing infrastructure. But I need to demystify a 5G rollout in Cambodia. It will not be anytime soon now a nationwide blanket coverage, right? I mean, nobody is expecting this. It's a clearly targeted urban hotspot approach. We are planning with roughly 150 or 200 sites for 2020, which is targeting exactly the hotspots where even with our relatively large chunk of spectrum, 4G comes to a level where we are capacity constrained. W e expect that the device ecosystem next year, mobile device ecosystem is expanding significantly.
I mean, iPhone will come with 5G in 2020, and we have in Smart 1.5 million iPhones in the network. There is demand for high-value phones. The Huaweis and Kodiaq are all coming out with many more phone models. We see a trend of Cambodian consumers to adopt relatively fast to high-end devices. That's why we believe in Smart's case, we see a case. I said it's not where we need to pump in now hundreds of millions of dollars of investments. It's a very targeted urban approach for the start. It's caveat to obviously getting from the regulator spectrum in C-band at an affordable price point. In Cambodia, we can do it because Cambodia doesn't have a satellite in orbit.
C-band is factually available for 5G, and industry and the government are very closely engaging on releasing the C-band somewhere in the first half of 2020 for 5G.
Yeah. So Adlan, you want to talk about without increasing data prices, how can you get your ROIC more than cost of capital other than the fact that you can bring down the cost of capital, right?
Yeah. I think here, I mean, we are able to do that if we are able to drive cost down faster than data yield. Data yield declining, right? I think it's always a race between cost and yield, right, and you look in time of a very competitive market, yield declined quite significantly, right?
However, in what we are trying to do now, in anticipation of potential tight competition and all that today, with Operational Excellence and how we build our framework, is to make sure that we have the lowest cost to serve in the market. Yeah? By doing so, I think even if, for example, if data prices are not, we are not able as an industry to increase data prices, we need to make sure that our cost per GB actually drops faster than the decline in data prices. And hence, that's why the focus on all this innovation, be it in looking at every line item of the cost structure to make sure how we make that changes, right? In the past, we have been looking a lot in a lot of the quick wins and all that.
But today, as part of our Operational Excellence, we are also looking at more structural changes, right? We have done quite well, I think, in terms of how we change the tower landscape in Indonesia. I think we probably need to do a lot more of that in order to realize and drive cost down even faster than what we have. We need to use a lot more digitalization. We need to use a lot more technology and a lot more automation, for example, to probably drive this. One of the key areas maybe in Indonesia is also we are quite reliant on the traditional dealers and all that, right? If we are able over a period of time to drive more towards the consumer, for example, there's quite significant savings that we probably can get.
So the answer is yes, if we are able to drive our cost per GB down faster than data yield dropping.
Yeah, there are a couple of questions to Thomas. I think three of them. Thomas, you want to take them?
Where can we start with the regulatory hunger to eventually take a piece of the growing profitability? I mean, so far, luckily, we have not seen any unpredicted move by the regulator. Besides that, with the introduction of the telecom law, USO and CBRD funds were instated, which then the industry has to feed into. But regulatory charges so far have been relatively predictable over the last years.
Yes, they have gone up, but in a way that we were aware when we invested into because those revenue shares which we have to pay to the government are mandated by the license that they are stepping up after certain years. So we had in 2018 one of these step-ups from 4%-7% of our revenue, but I said that is actually part of our license and predicted when we made investment decisions, we knew about it, right? T hat's okay, even though it's not without pain. The CBRD fund and the USO fund, 2% plus 1%, 2% USO, 1% CBRD, those came in 2016 with the telecom law. Then the industry has to feed in, which is nothing uncommon, and we can partially benefit from it again.
I would not consider it to now be in very high risk in Cambodia's context, but still, as I think in all our markets and not only in our markets, but worldwide, we need to work very intensely with the regulator to ensure that what regulators expect and taxation is coming in play is in balance with the benefits of our industry rolling into the market. Let me go to pricing. Well, indeed, very, very low pricing. Is there a market for premium offerings like content? Well, in a very limited scale, yes. But we are at the moment refraining from being too aggressive in that space. The Cambodian consumers are still very much into the mindset of free, free, free. Content has to be free. YouTube, Facebook offers obviously all this free content. We see a very limited willingness to pay for content as we speak.
But certainly, we are looking at it and exploring. We have particularly dedicated plans which are more from a free ride for premium video services from a data perspective or where we give a certain streaming balance. But for us as Smart to engage particularly in creating content or distributing content, we are at the moment not seeing yet the opportunity which we should really go into.
Okay. I think this is now the last question on Celcom brand. It's a very good question, actually. What is Celcom's brand position and what's the risk of destroying this position if you just go just in time? To me, Celcom is known for providing the best network experience and all that, right? So what's the risk of if you just go just in time?
I think we can do this properly if we would do it very, very carefully in terms of how we manage the just nice and just in time. Because as I explained earlier, we manage a few of the segments. We serve from Celcom's perspective, we serve all segments. We have the high-end, high-value customers. We also serve people who are the low-value customers who just spend 20 MYR, 30 MYR and just want to get whatever they can get on the network. So we look at this whole spectrum of customers. I think what's important that we look at here is a few things that we do in our planning. Number one is providing the level of service, the QoS in terms of level of service for the services that the customer wants, and also based on the subscribers and also based on the services that they actually subscribe.
I think that's the point of it. I t's not a general everybody gets the same kind of experience. Number two, we also look at in terms of benchmarking, in terms of area by area. We look at how is the performance against the competition, right? How we are into the area, whether cluster by cluster or even side by side or area by area by area. So we look at that benchmarking also. So the concept of just nice and just in time, and what we hope to do it properly, is not about destroying our value proposition or brand proposition, but actually trying to avoid in terms of overspending or over-investment in areas that we will not be able to use the capacity. Or in the end, we have what we call it underperforming sites or debt investment for what we call it, right?
By using, number one, in terms of giving the right service to the right people, the right QoS to the right services and the right customers, and benchmarking area by area against the competition, and also using the crowdsourcing data, I think we can target this properly and we can do this without risking our brand proposition value to our customers. This is something that we hope to do. We have Roy, our Chief Product, working at this together with CS there on the Chief Network, how the product and also the network delivery can actually deliver this without destroying the brand. I guess I hope that answered the question.
I think there's one last question which has come in. I think I'll take this as the last question, right?
For XL Axiata, given competitors, especially Indosat, are spending more on network investment, is there much pressure on XL Axiata to hike CapEx in Java and Ex-Java in 2020 to maintain average network in these regions to defend market share?
Yeah, I think for start, I think we probably need to bear in mind that Indosat have not spent for the last two years, right? So there's a lot of catching up that they probably need to do, which I think they probably started investing, especially in some of the strong areas in Java last year, and this year, I think they've strengthened it, and I think there is a plan for them to go outside Java. So far, we have not seen an aggressive rollout by Indosat outside Java, but I think with the money that they get from the tower sale, they will probably start expanding outside Java.
I think, is there pressure on us? I think as part of our framework that we shared just now, right? I mean, whatever we do in terms of our investment, we'll probably benchmark with the market leader, right? And typically, that will probably be Telkomsel. In terms of, let's say, aggressiveness in terms of rolling out in the existing area, probably not. But I think there will be some pressure coming from prices in some of these clusters where in the past we enjoyed a duopoly, for example, just with Telkomsel. But now there's a new competition coming in. But in any case, I think looking at there, there will be probably more competition, but I think the risk to XL would probably be not as great as what the market leader would probably see, right? Because we are a second-best alternate provider.
I think with our value proposition and momentum, we probably have some head start versus, let's say, in this case, it's Indosat. There will be some pressure on price, but I think if we continue to do what we're doing on a consistent basis, I think we should be able to continue to strive to get revenue growth in some of these areas outside Java, taking some market share, probably not as fast as what we saw in 2019. Nevertheless, I think that there will probably still be enough room for growth, let's say, looking at three players at least.
Okay, with this, I'll close. Thanks, Idham. Thanks, Adlan, and thank you, Thomas.
Thank you very much.
I'm going to give a bit of an introductory presentation like Vivek did earlier, and then I'm going to call my colleagues up so they don't have to sit here while I'm walking around. As Clare mentioned, Dominic's my name, and I'm going to be using this panel session to talk about how we remain customer-driven in the context of Operational Excellence. I think you've seen by now this morning with Tan Sri's presentation, as well as the last panel, that Operational Excellence is not just about cost. The customer remains at the core of everything we do. With Operational Excellence, there are five dimensions. I'm only going to talk about the second one, which is around optimum customer satisfaction and how we are achieving that and how we will continue to achieve that in the context of Operational Excellence.
I think the one thing I want to note is that our internal mission around Operational Excellence is, in one line, to be the lowest cost producer relative to our customer promise. T hat's to keep in mind the customer is always there, but to become the lowest cost producer, as Thomas saig before, we're a big factory. That's a large part of what we do. We have to deliver the lowest cost per bit. So achieving optimum customer satisfaction, what does that mean? And how is that different in terms of Operational Excellence? I think the first thing is we look at customer satisfaction as being a function of two influenceable outcomes. O ne of those is customer experience, and one of those is customer expectations. So customer satisfaction is not just about brute force investment.
There were a lot of questions before about, well, if you don't invest as much, does that dilute your level of customer satisfaction or customer experience? I think our answer is no, it doesn't. We've got examples both within our own operating companies in some markets, as well as we see other telcos around the world who have done similar things in terms of adjusting the customer experience and the customer expectation in order to achieve higher levels of customer satisfaction or at least maintain the same level while reducing cost to serve and reducing or maintaining CapEx as well. In terms of the customer expectation, what is that? Customer expectation, the reason I say we can influence it is because if you use a simple example, speed versus the video experience. A customer expectation is set by some market references.
When everyone's talking about speed, the expectation is about speed. When the conversation shifts to, "You want to watch video for longer," then the customer expectation is, "I want to watch video for longer." A s long as the quality is sufficient to what I expect, I'm very happy. I'm not going to measure the arbitrary speed that is coming. I'm watching it at the speed the video's playing. That's what matters to me. Yep. So that's a good example. It's personal. So segment of one analytics is very key. T his is why we say we can influence it. We have data, and I'll show you some later, about what our customers want from us and what they expect from us. A lso, we can influence some of the marketing messages around what the customers can expect from us and from our operating companies.
So it is something that we can influence. Secondly, around them delivering the experience. So the customer has an expectation. We need to deliver on that. That is all about relevance to that particular customer. So delivering a service, a video service, or delivering just a straight data service for browsing, whatever it may be, we can deliver those in a smarter way, different classes of service, fast lane, smart network, whatever we want to call it. And we can deliver those efficiently to deliver a certain level of customer experience. And it's relevant to that customer. So again, using analytics and understanding our customers, we know what's relevant to them when they want it. Utility, so does it do what it says? And consistency. It was talked about in the panel earlier that consistency is important.
The other thing our data tells us is in terms of customer loyalty, customer trust, and stickiness of the brand, consistent services do far more to drive stickiness to a brand and satisfaction than arbitrary or more random best experiences. I f one operator says, "I'm best," and then they don't do something very well, but I'm also best at this, and then not so well, but I'm best at this, they achieve less stickiness than an operator that provides consistent services, which is not necessarily best, obviously not worst, but is consistent and reliable. So these are the premises that we work on around customer satisfaction, expectations, and experience in Operational Excellence. So I mentioned here our promise is around the expectation to be the lowest cost producer relative to our customer promise.
We set a promise in the mind of our customers, whether that's around customer intimacy, product innovation, or price leadership, which are the three value disciplines that we follow externally. W e deliver on that through Operational Excellence in the most efficient way with a maximum consistency, using data to drive precision, precision for cross-selling, precision for upselling, and precision for knowing what the customer is expecting from us at the time we're going to deliver them a service. T hat's how we can differentiate in the experience hat gives us what we call that optimum customer satisfaction. I n the next slide, I'll talk about a bit of data on why it is we think this equation holds for us and why it makes sense that doing these things gives us optimum perceived value from the customer, fit for purpose, trust, and loyalty.
This is based on data. We do continuous brand health checking, customer perception surveys, both at group level, operating company level in all our markets on a quarterly basis, in some markets even more frequently. This is just a snapshot of what are the top four things in our markets that customers expect from us and value from us. I won't go into all the detail, but number one on the list is best perceived value for money. We say perceived because it's all based on expectation again. Value for money has different dimensions, but what it means for us is that we need to be the lowest cost producer to be able to offer a high-value, top-of-mind offer.
Now, of course, we try to use analytics to then upsell and cross-sell on the back of that high-value offer, but we need to be in a position where if a customer takes a headline low-value offer, we can still make money from it. And Smart has shown that that equation is true there. Secondly is around innovative brand equals premium. C ustomers tell us that we can have a low-cost mindset, but we should continue to be innovative because customers, particularly millennial customers, so those up to around 30 years old today, very much value innovation, and they perceive that as being a premium. It's not that we can necessarily charge a premium, but there's a perception of your brand being premium if you are innovative. So we don't lose sight of innovation. Operational Excellence still has innovation at the heart of it built in.
We are still saying we're going for profitable growth around home segment, digital value-added services, enterprise segment, as well as the consumer segment. Third is around the products and services that suit their lifestyle. So that's about relevance. It's about knowing the customer. K ey in here is segment of one, analytics. This is something we've talked about for years, if not a decade or so, around segment of one. This is something we do now regularly and of course w e'll talk about it. I think Allan has a couple of slides on it for XL in our panel discussion. We can and we do do segment of one. We know individually what our customers are seeking at a certain point in time and what's relevant to them. F ourthly, I talked about consistency.
It's just delivering consistently good experiences that don't have to shout and beat the chest about being the best. It's not about saying, "I have a 30 megabit per second service," but half the time it will burst up and down. It's about saying, "I will give you great user experience. Your video will start quickly, and it won't buffer, and it will look pretty good while you're watching it, and you're going to use less data, and you're going to be able to watch it for 30%, 40% longer than if you watched it on another network." T hat's around delivering what the customer wants in line with our promise. Now, how is that consistent with the five key areas of Operational Excellence? Well, best perceived value for money is about having a sustainable cost structure.
If we can be the lowest cost producer in terms of that bit factory component of our business, then we can pitch highest value, top-of-mind offers out there. If you think of it like a low-cost airline mindset, what you're pitched upfront is not what you pay in the end necessarily, but if a customer has flexibility to choose and add on, they generally will do that where it's relevant for them. Secondly is around innovative brand. So innovative brand is this sixth area here, which is we don't stop our operating companies from having their value discipline externally as they have done for the last few years, which is either product innovation, customer intimacy, or price leadership. That still remains at the heart.
However, Operational Excellence is what sets the boundaries of how we invest in those external value disciplines around being the most efficient, the most consistent, and being data-driven. Thirdly, is around the products and services that suit my lifestyle. Operational Excellence is not about reducing the portfolio of services we offer to customers. It's continuing to invest and deliver those services in consumer, home, enterprise, digital VAS, cross-selling third-party services through our platforms. T hat remains a core part of what we do to satisfy that innovation or premium aspect of our brands. So consistency of service, that's about having organizational excellence, sustainable cost structure, and that will enable us to deliver that consistency of service over and over again. T hat is sort of a hygiene factor that we look at. Finally, this is just a snapshot of what I mean when I talk about value discipline.
Over the past few years, all of our operating companies have chosen and have organized everything in their organizations along the lines of their chosen value discipline. T hat is either customer intimacy, product innovation, or price leadership. Now, that doesn't change. So it's not a major shift in strategy from that perspective. But in the middle, Operational Excellence is what both supports and sets the parameters for how the OpCos go forward in those external value disciplines. I n terms of innovation, that is still there. But of course, the way we invest in innovation is we become more efficient, and we can afford to then innovate in certain areas and invest in innovation. But of course, we don't want to innovate or deliver or develop different products and services just for the sake of it where they're not going to deliver a return.
So these are the slight subtleties that change, I guess, or shift under Operational Excellence. So that's hopefully an introduction to more detail you're going to get from our three panelists that I'll call up now. So I'll be joined by three colleagues. One is Allan Bonke, Chief Commercial Officer of XL. Please come up. Shanti, Chief Customer Service Experience Officer for Celcom, and Mahtab, CEO of Robi in Bangladesh. P lease welcome them up.
Hello. Good afternoon, everyone. I'll be taking on the couple of slides next to talk about how Celcom envisions to keep its customer promise in the context of the Operational Excellence. Okay. So customers obviously are very key to our vision. And in this sense, we actually strive to go beyond customer experience into customer intimacy in a very competitive market like we do here. I think customer loyalty and relationships are the key.
It is about creating evangelists and promoters to our brand. So far, our customer experience metrics are strong. We retain our number one position in relational NPS across the key operators. Complaints are at an all-time low. In terms of our touchpoint NPS as well, and especially on channels where we are managing face-to-face, the NPS is rated very high. Of course, in terms of our adoption towards digital, we're seeing 79% of our current transactions on a monthly basis are done on our digital platforms. The fastest growing platform for us is, of course, our Life app. It is actually managing about 2.6 million transactions right now. What are the key drivers in terms of the experience metrics that we saw? I think Idham spoke earlier around what we do around network experience.
We, in 2019 and 2018, also modernized by the end of the year, probably now 38 Blue Cubes. And this was more of a physical modernization and focusing around the senses in terms of the retail experience that we want to create. There's also a very focused drive around our call center to reduce costs and move towards more value transactions with the customers. And we've seen that reduction over the year now. A ll in all, having done this while managing the customer-related expenses, and it's going down. But of course, we're talking also about what we do in terms of customer experience around the context of Operational Excellence. W e believe this is really about delivering an optimum customer experience. T his can be done through simple and digital.
I think the way that the customers are also changing today put this very much in favor as our approach towards customer experience. S implicity in the processes, in the products that we offer, efficiency, I think being first-time right is so important as every kind of complaint that we have to manage is actually incurring costs to the company. So it's very important to be first-time right in our proposition and products that we offer. And of course, in the context of the consumers today and in the digital economy, being contextual and relevant to how we engage the customers remains a very important part of it. I'll go through the next few slides in terms of a bit more specifics of what we do in this area. In terms of retail experience, there are a couple of areas.
One is in the fact that retail as its purpose is also changing for consumers, and especially for telcos. Traditionally, we would look at it as a distribution presence, as a service touchpoint, but that is also changing in terms of what consumers expect. Optimization in terms of looking at what is the right footprint for Celcom today and in the next few years, and also what is the purpose that we want the stores to be. This will be different in certain geographical markets, for example. That is one part of it. Of course, profitability will be now the main driver in terms of measuring the performances of our stores. In-store experience is also important. This really affects our operating management of the stores.
So the more that we can have digitalization in terms of the engagement, a lot of self-serve, even in stores for simple inquiries, and how do we marry the adoption of our digital assets even in the physical stores as we use it to adopt more customers into our app? New business models are also something we're exploring, so in terms of running the stores, and today we are owning and running most of our Celcom Blue Cube stores, and this is an opportunity for us to explore some other ways to manage this. In terms of the Life app, the plan is to enhance what we call it is the 3.0, and we have reached about 1.8 million users now, and we intend to grow that because it really makes sense, right?
The more customers we have actually on the app, and that's how we can engage them on a one-to-one basis. And also then introduce what we call the personalized next best offer. I'll elaborate on this a bit in the next portion. And then we increase the value transaction that happens on the side. O f course, engagement is important. It's to get not only that they are checking the app once a month for how much I'm consuming or what my bills are. We want them to get into the app multiple times in a month. H ow do we do that? I t's important to create this engagement, loyalty benefits, exclusive offers, and likes. In terms of personalization, it's really around, that's where we say the crux of what customer intimacy is.
Basically, the more precise we segment and target our customers with contextual and relevant offers, the more conversion will improve. This is for us to manage churn, but as well as creating, increasing the lifetime value of our customers. So in terms of the approach towards segment of one, we are looking at omnichannel experience. So a customer would receive a specific offer to him or her on the app. But even when he or she walks into our Blue Cube stores, for example, the same offer will be presented to the customer. And of course, we try to use artificial intelligence in terms of creation of this next best offer, and of course, optimizing it to the right customers. L ast, I think even in terms of our contact center of the channel, I mean, people always ask, you know, do we still really need contact centers?
I think in terms of the context of the customer promise and expectations that Celcom has also set, it is still a channel for us. Of course, what we are zooming into now is making sure the contact center is also addressing more of the complex and high-value interactions, right? We try to actively deflect most of the simple inquiries and transactions into digital. With that, we expect further call reduction and also cost of the contact center. Also very important, every engagement, every interaction is an opportunity for us to cross-sell and upsell. Creating value through service is also a focus there. The third point is actually going back into optimum customer experience. Finding what is the right optimal customer experience metrics and doing this insights-based. We measure touchpoint NPS, we measure NPS.
We know what is that optimum threshold for a customer with regards to the experience that he or she, and we can. That's how we actually set our SLAs towards customer experience. Yeah, and so I've reached to the end of my slide. B asically, in terms of Operational Excellence, our objective will be to provide that optimum customer experience, still retaining the leadership in terms of our NPS and doing this driving towards digital, trying to get 60% of our smartphone users actually on the app. And of course, improvements and enhancement in terms of how our traditional channels create value for the future. Thank you very much.
Thanks, Shanti. I think, Allan, we have XL next. Yeah.
Thank you. Good afternoon. Just a couple of slides from XL regarding what we've been doing in 2019 and what we will be doing in 2020. I think I clicked this one. Yeah. A ctually, not to repeat what Pak Adlan said on the stage earlier, but he actually took away my punchline when talking about the revenue, right? So but I can brag one more. That's saying that, yes, we did pretty good revenue for the first nine months. I mean, even we're growing EBITDA faster than revenue. Now, looking at the subs. I just want to share with you, yes, we were growing the subs around 3% for the last nine months. Maybe it doesn't seem as significant as expected, but there's a couple of reasons for that. First of all, in Indonesia, the SIM registration process was introduced, and we are extremely strict when it comes to the ID from the customer when we do the acquisition.
But secondly, for the last 6 to 12 months, the main product in Indonesia has been the unlimited product. This has been introduced from two of our competitors. W e didn't jump that wagon, or we didn't jump that train and introduce that because we could not keep that customer promise if we introduced that unlimited product to our customer. So if you see the published numbers from our competitor, it didn't take off. There was absolute zero margin in this unlimited product. And now we actually managed to get one of our competitors to take out the unlimited product for the market. But then at the same time, when you look at the ARPU, we were actually able to work with our existing customers and our new customers to be able to increase the ARPU. I'll come back to that later.
Now, the two most asked questions I'm getting when we have this quarterly call after our published figures is mainly how's AXIS doing versus XL? And how's Ex-Java doing versus Java? And I'm not going to show you that. First of all, AXIS and XL are almost growing at the same speed. That's not a big difference. Yes, AXIS has been growing faster in 2019, mainly because we had some issues in 2018 when it comes to pricing for AXIS. So they're growing at the same pace. Actually, in Indonesia, AXIS is the fastest growing brand at the moment if we look away from Smartfren. Now, the next question is Ex-Java and Java. We know that Java is, of course, significantly bigger when it comes to revenue than Ex-Java.
But at the same time, with that investment, as Allan talked about, we have done for the last couple of years, we are growing eight times faster Ex-Java than we're doing in Java when it comes to revenue. W ith that, we have taken 3% market share in Ex-Java, and we are actually maintaining or defending whatever we call it in Java. We have the same market share. So with the investment we have done, we are growing in Ex-Java, and we actually maintain what we have in Java. Now, if some of you remember, I don't know how many of you were here in November 2017, I was standing on the stage talking about this left side of the slides, exactly the same product. So we have two brands, AXIS, XL.
We have three main segments, which is the white color, the blue color, and the young people of Indonesia. W e have three main hero products. And pretty unique in our industry, these three main hero products have been the same for more than two and a half years. So it's all about consistency, it's about simplicity. F or the first one, this is the white color, it's about the YouTube. You know that there's a lot of attachment towards the white color when we talk about YouTube. When we come to the blue colors, if you remember, I was standing two years ago with a fist in my hand saying, "It's human's rights to have free SMS and have free voice." And a lot of attachment for that for the blue colors.
The last one we have is the AXIS brand, which is mainly gathering the wallet of the youth and the gaming part of the youth. Now, how to measure are we successful or not when it comes to our branding, when it comes to the customer promise? Yes, of course, we can look at the EBITDA, we can look at the revenue. But here, if you can see, we're also looking at the NPS. And it's very clear when it comes to white collar, we are not number one. Telkomsel is number one in that space. But if you look at the last two years, we have been growing with a factor four, where the other guys have only been growing with a factor two.
We are closing the gap when it comes to the perception from the customer, the NPS, about the product and the brand that we have when it comes to the white collar. When it comes to the student, almost similar, but we have been number one for a couple of years now, even though we are actually growing at the same speed as our competitors. So meaning we are actually maintaining the gap. We actually even have a bigger gap from our competitors. I t seems like we are pretty successful when it comes to our target segments for these products. Now, what are we doing when it comes to XL? Just a couple of things. First of all, on the left side is about simplicity and consistency. On the middle side is about monetization.
On the right side is about, this is the most, I would say, exciting one, it's about digitalization. Now, on the left side, just a little bit of a heads-up regarding the YouTube. W hen we introduced the YouTube, we were not exactly sure what would happen. But now we've been running with that product for a while. You can see that for the customer who are buying the YouTube, you can see that the days of event, every time they do something, they buy a Waze service, they do the top-up, that has increased by 10 days. E ven that, looking at the traffic for the YouTuber, but not using YouTube, use some other traffic, that has increased by 16%. So it seems like the people who are buying this product, number one, there's a lot of attachment. We are close to the customer.
Secondly, they're going from the secondary SIM into the primary SIM having this product. At the same time, it's very difficult in a big country like Indonesia to be very granular with only three hero products. So what we have created is something called a voucher where you can actually be extremely granular down to each cluster, down to each city with different pricing and different top-ups for each of these products. It's very popular in the market as well. Now, on the middle side is about monetization. We have done a lot of things when it comes to that. But basically, just one thing, we introduced something called VIP. T hat just shows when you do something about your network and you have a product offering, it's actually attached to the customer.
We started out saying that if you want a VIP, you want a little bit better network as a customer in XL, you have to go into our digital space and do the registration. That we moved away from half a year ago, and now people have to pay IDR 10,000 to get that. With that, we can see now it's more than 14% of all the revenue coming from that product is actually on the VIP. C ustomers want to pay, they want to have a better priotization, they want to add something to the product, and they're ready to pay. Now, on the right side, this is about analytics. This is about CVM, CLM, customer management, or customer lifecycle management. I just want to give you two examples. W hat we have introduced is XL is something called NBO, the next best offer.
Today, we have around 15 channels in XL: retailer, our app, our web, our modern trade channel, etc. 15 different channels. We have created with this mature, what we call it, analytics tool that we have today, 900,000 real-time profiles of our customers, meaning that we can personalize up to 900,000 different customers. So for each customer today, we have 20 different products. When they go into one channel, it can be a retailer, they'll have exactly the same product offerings as they will go into the modern trade or they will go into our channel or apps or whatever. Exactly the same offering they will have all around. Secondly, what we have tried and what we are doing a little bit of a test is a non-telco, so today, yes, we have 55 million customers. Every day, we have 10 million hits on USSD.
10 million customers asking us about what is our balance. Star, one, two, three, we are getting a USSD, and we are sending a menu back to our customer. Six different line items. Now, here's your balance or buy something or do something else. I magine that it's 10 million hits every day who want to talk to us. So what we have tried is, guys, we're trying to do some non-telco product. O ne thing was actually McDonald's. So we had a line item. If you want to buy McDonald's, you can have X% of discount. Some of the days in Indonesia, we are the biggest distributor of McDonald's. Secondly, we're also trying with insurance and handset insurance, etc. So usually, when you do this in Facebook, you have something called CTR, click-through rate. It's around just below 1%.
With this, doing it in our own channel, we have around 2.9%, three times as higher click-through rate that people are actually clicking and want to buy. That's pretty significant. Now, for AXIS, again, one of the most common questions we are getting with this quarterly call is, how's AXIS doing? What about the pricing? What's happening with the yield, etc.? W e know that AXIS, the promise is price leadership. Being a price leader doesn't mean you're the cheapest in the market. Even though you want to be extremely competitive, it does not mean there's no yield on the product. W hat we have done with AXIS, we have basically introduced something called sachet. This is low-validity product, one day, three days, five days. O n that, you can actually boost your subscription by having one day YouTube, one day Facebook, one day WhatsApp, etc.
You can create your own product with very short validity. In that way, we are together in the youth wallet, which is limited, and at the same time, they can buy the gaming which they like so much. As you can see on this graph, it's been a pretty significant success story that we have had for the same January. Even looking at the yield, the yield for this particular product is better than some of the yield that we are seeing from our competitors for the published and non-published figures. That's actually pretty significant. Now, on the last slide here, I just want to tell you what are we doing going forward. We have to pick our battles here.
So the two battles we want to pick is basically the digital space, our ecosystem when it comes to our app, and then it's about the analytics. Now, we cannot just have an app which is equal to anything else we're doing in the market. W hat we have decided, the app's going to be unique, meaning that when you go into the app space, you can find product that you cannot find anywhere else. At the same time, you will have unique services, unique offerings in that world of our own app. Example is like, yes, if you want to have a limit on your pay as you go, you can put a limit within the app, meaning that you will get an alert every time you reach that limit. We will also have a lot of kind of, what you say, emergency balance. I
f you run out of balance, we can actually lend you to that emergency balance, which you only can do within the apps. So we're going to develop that app mainly to be close to our customer to make personalized and tailor-made offerings. Secondly, with the investment that we have done in analytics, I think we are way ahead of some of our competitors. Very, very clear. We are pretty mature when it comes to that. So here, I'll just write down a lot of things that we are doing. But just to give you an example. So today, if a customer goes into a retailer, he will ask to buy something, and the retailer will start punching in his number. When we get that number in our base, we actually immediately send back a new offer, both to the retailer and to the customer.
Before he buys, he can actually say, wow, if I just buy 2GB more, I can get that for IDR 5,000, and I can see whatever it is. I'm better at the WhatsApp or YouTube, whatever I'm using. We can tailor-made that offering both to the retailer and to the customer. The retailer will get some benefit by sitting in that product, and the customer as well will get some benefit. At the end of the day, we, of course, want to have that transaction into our own app, into our digital app. It's pretty advanced. We have invested a lot of money, both in tools, systems, and people, but we are ahead of our competitor, and we are pretty proud of that. That's basically what we're doing, and I assume there will be some questions regarding this later on.
Thank you, Allan. I just want to remind everyone, Mahtab is going to step up now and give his presentation. Then we're going to go to Q&A. There are roving mics, so if you don't want to type the question into the app, please put your hand up, and then you can have a mic and ask the question directly. Thanks. Mahtab. Yeah.
Good afternoon, everyone. I think you already know about Robi performance for the last three years. 2016 was pre-merger, was one of the worst years, and then we have turned around in the last three years. Just to give you an overall number, in three years, our revenue growth was almost a high double-digit CAGR. If you look at EBITDA growth, it was 4x of revenue growth. That kind of performance we have shared.
You know that this kind of performance was possible for a couple of strategies which worked exceedingly well for us through its execution. One of that was dual-brand strategy. The second one, I would say, [Uncertain] strategy. Third one is 4G leadership strategy. Overall of that was Operational Excellence. That's what I'm going to share with you, primarily focus on dual-brand strategy, how it really helped us to excel in the last three years. Now, if you look at, we have got two brands. I think there are very few successful two-brand strategies globally available. We can strongly claim now Robi has two dual-brand strategies which are successfully moving side by side. If you look at Robi, Robi is an innovative digital brand. In case of Airtel, it is very clearly focused on price leadership and focusing on youth segment.
These two are clearly identifiable. If you look at the next slide, what I'm going to show you is it's complementing each other. It's not fighting with each other. Therefore, we are getting the best of these two brands. How? If you look at Robi, the pricing strategy is very clear. Robi is targeting to the market leader who is considered to be the premium brand. So Robi pricing in most of the markets is either at par with GP or slightly lower than GP. I f you look at the ARPU, ARPD rate, you'll find that true. I n case of Airtel, another price brand that we have in our country, which is a third player, which is Banglalink. And there we are keeping in Airtel's strong markets at par or higher than Banglalink, or in other markets, it is lower than Banglalink.
In case of Geostrategy, I think that's where the beauty comes. In case of Robi, it's focusing aggressively on [Uncertain]. That's the strategy I just talked about, which has been working very well for us and a little bit in Dhaka. In case of Airtel, we are focusing on [Uncertain] plus Dhaka is not so much focused. It's a low focus because they are one of the key players in the Dhaka market. The strategy is very distinct and clear, as you can see from here. Now let's look at the number. By driving this strategy, what kind of achievement do we have got? It's the 11 quarters of successful growth. You can say this kind of growth is sustainable only when it is sustainable for this kind of 11 quarters.
If we're starting from quarter one, just immediately after merger, actually, I would say it was during the merger process. Customer market share was 27.6%, moved up to 29.5%. Service revenue market share went up by 2.7 percentage points. M ore importantly, data revenue market share went up by 5.4 percentage points. And you may ask why data is growing so fast compared to others. One of the reasons why Robi has been very clear in terms of investment strategy, that is to focus on data. Th at's why we tried to take the leadership of 4G at the time of 4G launch last year. Unlike our number one player, GP, who is still focusing on 2G and 3G investment. T hat's where you find the differentiation.
But I think the Robi strategy of focusing on data and taking a lead on that is the right thing to do, and that is being reflected here. evenue growth, quarter on quarter, if you look at over 11 quarters, it has gone up by 31.5%, and data revenue is 108%. Again, it shows the number. I n case of Net Promoter Score, actually, till two quarters back, we were number one and number two brand in terms of Net Promoter Score. But because of our 2019 bit of profit focus, we have lost a bit, but it was just marginally. D ual-brand play, if y ou look at both the brands, are gaining from the competition. T hat's very interesting, and we'll show it later.
If you look at from CCD on the right side, our game plan from CCD to national play, there you can see that [Uncertain] used to have a contribution of 16.1%, and now that has gone up to 21.5%. It means around 5.4 percentage points gaining in the [Uncertain] market. It's not only in terms of revenue. Even if you look at EBITDA, ROIC, [Uncertain] remain one of the critical contributors to our growth of the last three years. If you look at 4G users, we have been growing the fastest. Even today, our 4G market share remains around 48%-49%. If you look at data volume, 4G data volume contributes 46% of the total. More importantly, last year during August, that MNP, Mobile Number Portability, launched. If you see, we are growing far faster than the competition.
In case of one-to-four port-out ratio, if you look at quarter three, it was 8x, and now, as of now, if you look at, it is around 4.1x, so in port-out, all the numbers, if you look at, the two-brand strategy is contributing and reflected in all these growth numbers, as you can see here. Now, another thing that I said that Operational Excellence remains at the heart of whatever Robi does. Without Operational Excellence, we cannot turn around the business as we have been talking, and if you look at EBITDA of two brands, and we track two brands' profitability separately, and when we started, EBITDA of Airtel was pathetic, and the cost structure was also pathetic, and we had to fix it, and if you look at, in case of Robi, the EBITDA profitability went up from 2.25 to 5.42, almost 2.4x.
In case of Airtel, it was meager 0.11 billion to 1.91 billion, which is 17x. A good chunk of that is also driven by the merger synergy that we obtained in the year 2017 and 2018. But more importantly, if you look at Airtel as a price leadership brand, the price of that is almost 25%-30% lower than or cheaper than even Robi. Despite that, if you look at the EBITDA margin, it remains healthy and is very close to Robi. When we started, the gap was almost 15 percentage points, and now we have brought it down to almost 4 percentage points. We have a vision, a clear vision to bring it to the same level. If you look at contribution to incremental EBITDA, you will see that Airtel is contributing around 36% and Robi is 64%. So two-thirds and one-third.
E ven from operational efficiency perspective and profitability perspective, both the brands are contributing heavily to our journey. Let's look at that one component. Network, already Vivek has mentioned in his slide how we are maintaining network costs, so I'm not dwelling on that network cost, but one thing I can tell you, last three years, in absolute terms, Robi's, despite significant growth in revenue, significant growth in network, but our absolute costs remain at the same level, and just one glimpse of marketing costs that we are sharing here. In case of Robi, the cost has come down from 25.2% to 20.2%. In case of Airtel, 44.5% cost has come down to 24.2%, so overall, the decline is 7.8 percentage points, and that's the ultimate example of two-brand strategy along with Operational Excellence, and that's the last slide from Robi.
And here, what we are articulating is, yes, initially, the first two years, we focused on growth strategy, moderate growth strategy. But in 2019, we started focusing more on cash and margin. And that strategy will continue even in 2020 and onwards. T he key milestone remains on, we'll continue to grow above the market. We'll continue with dual-brand strategy. O f course, we have got a different game plan to play differently to gain more. C ost excellence will remain at the heart of Robi's way of doing things. Value discipline, as we have clearly said, Robi is competing with the number one brand in the country, and it will be based on innovations. We have got an innovation roadmap. Every quarter, we bring something new. I'm happy to share that with you.
Last three years, whatever new technology, whether it's 4G or VoLTE, whatever comes out, Robi remained number one in terms of bringing new technology in the market. And price leadership, if you look at, Airtel will remain a price leader, and that game plan will continue. But the only challenge we have is to improve the profitability of Airtel a bit to bring it to the same level as Robi. I t will be powered by digitization, analytics, and AI, and more importantly, cost re-engineering. The current way of doing things will not take us too far. So we have to change the game plan to bring more cost benefits for the operations. I n summary, I think our dual-brand strategy is working extremely well on the proposition of innovation and price leadership for two brands. T hat's it from my side. Thank you very much.
Thanks, Mahtab. We're into our Q&A phase. I've already seen three questions on the app, so I think I'll turn to those first, probably starti ng with Sultan. I think, Shanti, you've had a chance to see that first question. I'll just repeat. What's covered under relational NPS statistics? Is it all collated? Does it include MCMC and social reputation scoring, or just based on internal monitoring? I think that's the first question.
Okay. Basically, our NPS is a relational NPS. It is an independent study done by Nielsen. What happens is actually the customers of Celcom, Digi, Maxis, and even U Mobile are sort of asked a set of survey questions. These questions cover all aspects of experience, primarily network, product, and customer service, as well as some of the other value-added services that all of us provide.
The customers actually rate the providers, their providers, by the NPS methodology, and that's how the scores are derived. On the second question, second part. Okay. The second part of it is in terms of right-sizing recruitment. Today for Celcom, Blue Cube, as well as our retail channel partners, we have more than 240 physical presences around the country, and out of that, 58 is actually what you see Blue Cube, and that's basically Celcom own and operate. The idea of this optimization is to look at which areas do we need to be continuously present and for what purpose. There are some areas where it's still a distribution presence transaction, and we see the most efficient to do this is by our retail partners.
Definitely from a brand engagement, there are certain stores that we need to keep for our interactive experiential, especially in the key market centers. So overall, this optimization is to make it fit for purpose. Yeah, it's a question of there may not be all the stores that Celcom needs to operate in the future as well.
Thanks, Shanti. Second question, I think also has two parts. Mahtab, do you want to tackle that?
Okay. I think the first bit is very clear whether it is sustainable. As far as we are concerned, it's pretty much sustainable, and rather we can further improve on that. T hat is one part is very clear. The second part is regulatory situation. Currently, if you are referring to the current situation, it is in the court.
Even if it is in the court from Robi perspective, I would say it is not going to have a severe impact on Robi as such because we feel the merit of the case is so strong. We should be able to eventually be able to handle it. In terms of future regulatory situation, it's difficult to predict, but consider the current regulatory situation, we may say it is manageable.
The third question here, I'm just reading. I wonder whether maybe, Mahtab, you can comment on that second part of this question because you mentioned in your presentation that there was a slight reduction in NPS this year, and you attributed that to, I guess, pulling back in some areas. So can you maybe explain that first, and then maybe Shanti can answer the first part relative to Celcom?
Okay. As I said, if you look at one of the objectives we have had was Robi and the Airtel brand, despite having different price proposition, still need to have a similar level of profitability. And to do that, we had to increase the prices of Airtel in certain markets. T hat has helped us to bridge that gap. But the main score of NPS for Airtel was coming from pricing. The ment you take that price increase, it is bound to have a bit of impact on pricing, price NPS. I t was a very conscious call that we have taken. T he good thing is the gap is not so significant for which we should be worried about.
Thanks. Shanti, you want to relate that to the Celcom situation?
Yeah, it is. Of course, it's a balance like walking on a tightrope.
But I think this relates to how we can be insights-driven. So basically, with the NPS data and the service levels that we have set today, we are able to actually find an optimum level of service that we can provide. So I give maybe an easy, relatable example, right? Let's say a contact center, and there is a service level of 80% of the call picked up in 20 seconds. And we have a certain level of NPS that we have locked in for that. T he data can actually tell us whether we need to be 80/20 or not, and yet can command that NPS level. T hese are some of the metrics that we are looking at, not just around the customer service touchpoints, but also network in itself, as we spoke about just now, video experience, what's the right service level to set for that.
But of course, a caveat here. I think for Malaysia, we have certain mandatory standards, quality of service that is being set by MCMC. T hat is kind of like the minimum we have to meet. But everything above that should be insights-driven, should be to what the customers are telling us, what is the optimum level that's making them happy.
I guess it comes down to knowing what our customers want from us. I think that way we can be just in time, just before or just when they need it. I just noticed the next question that came up is a little off topic, but I guess does Vivek or Tan Sri Jamal want to make a comment on that question? Not really for our panel.
Okay. How do I answer this question? So if we increase our, I guess if they put on us, we will, right? I think that's quite mechanical in that sense. The question is, will that happen? I can't speak on behalf of VoLTE for sure. It is their option to exercise. Actually, it's not us. In theory, it will have to happen or they have a chance to do it by end of this year, by end of 2020. You should ask Mahtab, does he prefer which way?
We might leave that for the break conversation and then turn to the topic. I think we still have a few minutes left here. There are roving mics if you want to answer questions. There are no more in the app right now, so I'm going to ask a couple of questions. So maybe one to Shanti.
For Celcom, as we've seen, customer intimacy is your value discipline and effectively your promise. Can you give us maybe another example or two about how you're achieving high touch while still keeping cost to serve low? Are you doing that through high tech or other means? Give us maybe an example or two.
Yeah, a couple of approaches to that. The crux of it is a bit of what I mentioned just now. Basically, we focus to drive more adoption of our app as sort of the key interaction and engagement points for our customers. Using technology, but still remaining high touch where we can communicate personalized next best offers and ensure that in terms of support, the customers are also able to get it there. Any kind of service that they want to add on is also available in that.
So it is basically pushing more of that. We've also started sort of pushing a lot, so the prepaid market in Malaysia is also rationalizing and consolidating. And so pushing a lot of customers converting from prepaid to postpaid also is a way for us to ensure that we continue to achieve that sort of stronger stickiness of customer engagement and loyalty with us. T he se are some of the initiatives for that.
Okay, thank you. I'm going to turn to Allan XL now. So Allan, I think you mentioned quite a bit about analytics and how you are getting close to the customer. You've got, I guess, two brands, so you've got different value disciplines. But maybe explain to us a little bit more about how you're using analytics to drive understanding what the customer wants and increasing your ARPU.
First of all, we have an ambition to be number one in our market when it comes to analytics, right? And the reason being, first of all, in the spirit of Operational Excellence, it's a very cost-effective way of touching your customer, right? S econdly, we strongly believe that personalization and being close to your customer, communicate one-to-one is much stronger than any other customer promise that you can have. The big question, of course, is where are you on this maturity ladder where we know the results? We all know that, yes, we are able to talk about the past. We can analyze on the past. I think everybody can do that. We're even able, by doing that past analysis, we can actually predict about the future.
But we're also in a stage right now where we can do prescriptive, where we can say, yes, we know what happened in the past. We know what's going to happen in the future. But at the same time, we also know when it's going to happen and what kind of task we're going to introduce to avoid it or actually to increase it. O f course, the last mile is going to be, as Tan Sri talked about, it's going to be the AI, the cognitive, right? We are not there yet, but we are doing the investment, and we are doing the investment in both people and system to get to that stage. I would say it's a complex thing to do, but we are not big time, but at least we are ahead of our competitor.
By that, as you saw some of the tasks that we're doing, we are actually able to interact with the customer in a much different way. We know that even though we have been running two and a half years with these three hero products, it's not going to continue. We need to create personalized offers for customers who are good at WhatsApp or YouTube, whatever, and they know exactly how many GB they need. I t can be one-to-one communication, as many customers we have, as many products. Without customers seeing it, it's a complex thing.
Very good. I think we've got a couple more questions here. First two for Shanti again.
In terms of social media presence, so Celcom, we have Facebook, Instagram, Twitter, and we do monitor our presence there. We actually have a social media hub, which is a team that operates 24/7.
They are actually divided into three groups, basically what we call the brand builders, also the service aspect of social. W e do the service element also basically. So as a first filter, we have a chatbot, and then they do sort of the onward engagement. O f course, the third one, we're also looking into how we use social media to sort of generate leads for upsell and cross-sell. So basically, we have a social listening tool. That's how we also monitor sentiments and basically engagement, share of the market vis-à-vis our competition as well.
The second question about types of questions into the box.
Y es. Yeah.
Basically, in terms of, we've done a lot of initiatives actually this year to kind of also do a bit of proactive management before sort of complaints get in, especially when it comes to network outages and all of that, more visibility to customers. The calls that come in today, we still have about more than 50% of our calls are related to inquiries and some simple transactions. These are the ones that I mentioned just now that we need to deflect as much as possible. As a stance, we've been quite soft in our deflection, but increasingly, we are getting a little bit more focused around really the simple transactions we can basically direct the customer to the app, for example. In terms of cost of solving the issue, yes, very much.
I think every complaint that gets generated as a ticket and having to be resolved throughout the chain within the company will cost the company more money than actually the call itself, which is why the proactive management in terms of how we can actually reduce even before a complaint comes and using all the tools such as IVR as deflection is very important.
XL. I think Allan, Read the question and answer.
Yes. How are your competitors marketing? I know exactly how the marketing story is, right? That's to copy us. So they are basically copying us in everything we're doing. W hen we introduced YouTube, now everybody introduced YouTube, the free YouTube. When we introduced free SMS, free online, everybody introduced free online. T he last one surprised me a lot. When we introduced the sachet, one day, three days, five days, now Telkomsel has introduced that as well.
For me, that's crazy, right? They have a base of how many customers, four, five times bigger than us. Now existing customers can go in and buy one, two, three, or five days' sachet product to a much, much lower price. That can hit them very badly. In the market, not that we are much better than the other guys, but actually they are copying what we are doing. Now, when it comes to Ex-Java and Java, yes, it varies the strategy that we have depending on are we number two entering the market, are we number three or number four entering the market in that space? Secondly, do we defend or do we attack? Not to reveal what we exactly are doing, but we have different go-to-market strategies depending on where we are and where we are going in.
Thanks, Allan. Mahtab, I think maybe what might be the final question for Robi. Okay. I think, yes, there is a further opportunity to reduce cost if I take out the regulatory element of it. The focus on cost, yes, there is an opportunity. But we have to look at it from a different perspective. If we go aggressively on non-voice strategy, which we have not been so far, we are going selectively. If we go aggressively, there may be short-term pain, but long-term, medium to long-term, there is a good improvement in EBITDA. If we ignore the non-voice strategy, I'm talking about beyond moderate, in that case, there will be some improvement in EBITDA in the coming years. Should be, as I touch wood keeping my fingers crossed.
Thanks, Mahtab. Last one, I think, for XL.
So the journey has just started regarding the apps. I'm not talking about how many we have. We do not have enough. That's very clear. There has been a lot of challenges in that app. First of all, if you look at the Indonesian market at the moment, this is a super apps market. I think you all have seen that. This is the XL, this is the OVO, the Ava, whatever they're called, and all of them are significantly big. They're not monetizing what they're doing. They're just talking about how many clicks do they have every day. That's what we're up against. At the same time, what we have now introduced or had a trial that is going to be a unique channel, not a channel as we have all around, but actually, as I said before, where we can get unique products, unique services. With that, we can see a significant increase.
That started a couple of months ago when it comes to the AXIS brand. We really can see that if you want to buy something special, you have to go into AXISnet. We can see the increase going up. Our ambition is, yes, it will be 30%-40% for the next three to four years coming from our own apps. Not saying that it's our own apps in the app, but it could also happen that we have partnerships where we have our ecosystem within some of these super apps. Very ambitious for the next three years, but we need to develop that channel as it's much more cost-efficient. We can talk one-to-one, communicate one-to-one with our customers through that channel.
I did say last question, but I think there is. Is that a question you can answer here or not? If not, that's okay.
No, that I cannot.
I thought so. That's okay. I think we're done with time. So can I thank my guests and give them a warm round of applause? Thank you.
Okay. I will be talking to you a little bit about the digitization efforts, the context framework, and also some of the results as we transition from an implementation and execution phase to monetization. Axiata has, as you might recall from previous presentations, Axiata has consistently adopted a five-core framework for digitization, starting with on the periphery, the product and service layer. Product and services in the digital domain would mean, for example, the suite of media services or the digital engagement that is provided to various value-added services, what we would call digital value-added services. These are consumption of an augmented set of mobile services using digital interaction mediums.
The next layer is on external interfaces. This is how we engage with our customers, both enterprise as well as retail consumers. The app that we have been discussing at great length even in the previous session is a good example. The websites, the various digital engagement channels provided for customers to engage with the service desk as well as with various parts of the operation of the company. Internal processes, the real heart of the organization, ranging from finance processes to HR to operational processes, and also the interface between the front-end processes and front-end systems, such as, for example, a dealer management system and the back-end finance. Infrastructure and platforms is the fourth core. T his is a very important core. On one hand, it encompasses the entire network layer, which I will come to towards the end of the presentation.
I won't spend time on that right now, the entire infrastructure, but also the platforms which interface the infrastructure with the front-end systems. So our billing systems, our customer relationship management systems, the API platforms, etc., which provide that bridge between the hard infrastructure and the front-end systems. Last but not least, organization and culture, the very heart of the people and the importance of digitizing and changing the mindset of the Axiata family across all OpCos, across all operating markets to enable the people to work in the environment of a digitized ecosystem. T hese five cores have assisted us in disciplining activities, giving meaning to the roadmap and blueprint that we have adopted over the last three years. As you can see from this, one key element or feature of this is that it's all-encompassing.
When Tan Sri Jamal hit the reset button on digitization, an important challenge given to the team was to make it holistic, make the entire organization digital, and that also posed the challenge of having to do this in flight because unlike the digital startup telcos who start with nothing, no customers, no infrastructure, no balance sheet, basically, for us, we have millions of customers, millions of expectations because customers have already established expectations from their telcos, infrastructure, legacy systems, some of which are aging, and then the challenge of digitizing end-to-end in flight. So this, on one hand, is a positive, having all the legacy and infrastructure because you have the strength of being able to deliver under your own control without having to rely on others to carry your traffic, but on the other hand, it's challenging as well. Now, why do you digitize?
On one hand, it's for operational efficiency, speed, customer intimacy, flexibility, agility, etc. On the other hand, why you digitize is to produce large amounts of big data. That is the less spoken of outcome of digitization because if you digitize your organization end-to-end at every single interface, at every single transaction, the big data pool gets augmented. W hile if we were to implement, for example, a dealer management system today and we digitize the dealer process, a transaction which never got recorded in the big data lake tomorrow gets recorded. A s this data lake gets augmented, it provides a very rich resource for analytics and AI. As you all know, AI would be impotent if not for the data.
While the technologies are all very sexy and powerful, if you do not generate increasing amounts of data and deep data at that, the power of analytics gets muted. So combine these two and also combine it with capability augmentation and IT transformation. IT transformation is the platform layer that I mentioned earlier. T hat is actually breaking up your IT systems to sing and dance at the speed of digital. Second is to have your people singing and dancing at the speed of this digital world. So these are two augmentations which had to go in parallel. And together, we aim to deliver functional excellence, Operational Exellence, and competitive advantage. The question that we have now begun to ask ourselves since we have moved considerably down the execution path, can we achieve digitized telco economics? T hat is the challenge the organization has posed itself today.
Can we match the best in the world in terms of the economics enjoyed by a digitized telco? That's the framework, and I'll just whiz through a few proof points. To begin with, during the execution phase, we paced ourselves against external benchmarks, in this case, the Oliver Wyman Digital Telco Index, pacing ourselves against other telcos globally in terms of digital maturity across some 65 KPIs, both at the front end as well as the business process internally. Over the years, as you can see here, our OpCos have progressed very aggressively, and one or two of them have got very, very close to the global telco top quartile median. Now, that means that if the speed of change and the speed of improvement continues, that two or three of our telcos will move to be globally best and exceed the global top quartile median.
In 2019, and by the way, the Digital Telco Index continues. We are just moving into the phase of the 2019 evaluation at the end of this year. During this current year of 2019, we also took on the challenge of evaluating our digital maturity at a functional level because the tendency is that you tend to charge the responsibility of digitization with the chief digital officer. It becomes a functionally siloed effort, and in 2019, we took on the challenge. We made every CXO responsible for their own digitization agenda and measured maturity against a global scale carried out again by a third party, Frost & Sullivan. T he movement that we saw during the course of the year was very encouraging, and we will continue to push the boundaries on functional digitization as well.
Again, overall, our aspiration is to take three of our OpCos into the global top quartile and also to ensure that there's deep digitization maturity uniformly across all functions. During 2019, the LEAP program, which was actually the Frost & Sullivan-based functional maturity exercise, spawned over 250 unique initiatives. I don't intend to take you through all of these, but basically, it's a proof point that the 250 are spread across the five cores of digitization and also the functional domains of our operations. J ust to illustrate two of these just to give you the context and a feel for what a digitization initiative might be. But just keep in mind that these are just two out of 250 across the group. At XL and Celcom, there are significant gains from process automation in finance and customer experience. I think Shanti spoke about some of this already.
Similarly, in finance, where bots are used to automate processes and also automate the interface between the offline and the online, these have resulted in doubling the speed of turnaround time, both for internal as well as external processes. The first point of contact now becomes digital, and thereby alleviating the requirement for manual process at the frontline, errors creeping in at the first point of contact, which is a very powerful reason for including digital at the first point of contact, avoid garbage in because you get the first input accurate, and also a seamless handover from bot to agent as the complexity of the interaction increases.
Likewise, if you look at dealer management, which is, as you know, in emerging markets, especially one of the most complex processes, you're talking about reaching a customer thousands of miles in some cases or hundreds of miles in a rural area of the country and enabling the activation of a SIM card, performing a KYC, obtaining the customer details, and transmitting those details into the core systems, activating that customer with 100% accuracy. This is no mean task and translates to a lot of cost in the core operation of customer acquisition. Now, what has been done in several of our OpCos, and what you see here is an example of the Dialog Retail Hub. This device sits at thousands of retail points across the country, and likewise, the Robi trade marketing app and the Smart dealer management app. The results here have been phenomenal.
A 45-day process has been reduced to a couple of minutes and, in the best case, one minute because a customer can now walk up to the most remote edge of the customer service network. At that point, just hand over documents, thumbprint, get a photograph, sign digitally, and the documents are transmitted paperless all the way to the center, activating a customer with full KYC and regulatory compliance. In Sri Lanka, 12 million A4 pages saved annually. The Sri Lanka example provides around $1 million annual savings. This is just for one initiative. This, I hope, gives you an idea of the type of initiatives that have been driven with a wide functional dispersion so that you get similar across the other functions and across the various cores of digitization.
Now, that sounds exciting, and you've got a feel for the numbers, but I think that's not good enough. We need to build a model which translates digital maturity in a very process-oriented way from maturity to financial outcomes and therefore to a best-in-class cost structure. This is not easy to do. If you were to challenge any telco, I would say globally, I would chance a guess that even globally, they do not have a firm correlation between their digitization effort and the financial outcome. Now, we strongly believe that this is strongly correlated. It's all about establishing it so that then we can reduce or increase our focus on the right initiatives at the right time to get Operational Excellence delivered in the shortest possible time.
So what we have done here is, together with our partner, A.T. Kearney, set out to identify digital leaders globally, telcos who are digital leaders, then to look at the P&L of these digital leaders, the cost drivers, and also to identify the top KPIs per functional area. H ere we are talking about 100, 150 cost drivers, breaking the P&L down into a very detailed segmented logic, financial logic, reaching the various margin layers of a P&L. Post this, quantified impact of the top KPIs by each area, what we call cost curve. T hese cost curves have to be normalized to operating environments. T he cost curve for Indonesia for customer acquisition would be very different to the cost curve for servicing a customer in Sri Lanka, for example.
So the cost curves are developed for each activity by market and thereby build a target P&L for each OPCO with a three- to five-year horizon. Now, the ambition here is to achieve that digitized telco P&L, which is under construction within that three- to five-year horizon. W e could envisage, looking at the digital leaders today, that there is a margin arbitrage of somewhere in the range of 3%-8%. I'm giving a wide range because execution efficiencies vary market by market. But this is something which can transform the P&L in real terms through digitization across all functions. O f course, from block four to five, it's a translation of these high-level P&L models into specific initiatives, enabling tracking of the initiative as well as the convergence towards the target P&L over time. Moving on to analytics, similar good news on the analytics maturity.
Here we refer to the TM Forum benchmarks methodology as well as external evaluations, and our maturity in terms of analytics as a practice, the use of analytics, and the embedding of analytics in business processes over the period of mid-2018 up to December is very aggressive. As you can see, our target is to get to 61% maturity group-wide. The 70%-80% watermark is what one could call global best practice. Y ou could see that several of our OpCos would hit and exceed this level in the near future. The Axiata Analytics Group, which started off with around 60 analytics professionals two years ago, has now been augmented both through acquisition of new talent, but interestingly, through training to a large extent, retraining as well as applied training across the group.
Today, we have a very large pool of analytics professionals, 270 to be exact, across the Axiata Group. The financial returns from analytics, a good example is the Next Best Action chatbots, fraud management as functional examples of where financial outcomes, Operational Excellence outcomes have been derived. This table shows you a spread of analytical models built across the functional verticals and also their degree of replication because a model which is developed for one market could, with minor adaptation, be translated into another market and a different environment. We have an example here of where the Next Best Action model was translated from XL to Celcom and also to Robi. Dealer optimization in XL and Smart have been carried out almost concurrently.
Again, these are just a few examples to enlighten you on the focus on the use of analytics, but also replication, building of models, and using the efficiencies of common data models across the group to learn from one market and replicate across others. T here are two examples that were given. We take the Next Best Action model. We started at Dialog, moved to XL, improved take-up between the Nepalese market and Sri Lanka. A newer version of it was deployed in Indonesia and then brought into Malaysia, the results being far improved. Now, the same model will be recycled back to the first two OpCos in Dialog and XL to achieve the same kind of results. An interesting example on dealer management, the dealer optimization model in Cambodia and Indonesia.
Here, it's about identifying the demand parameters as well as the supply and efficiency parameters of the dealer ecosystems in order to decide where a new dealer is required and when and with what terms and conditions those dealers should be appointed to get the financial outcome that is targeted. Add to that front-end digitization like apps and mobile dealer management interface to complete the cycle and the monetization of the initiative. All this comes with the challenge of building capabilities and therefore the parallel focus on both DevOps and analytics. We took a firm decision around two years ago with the initiation of the IT modernization roadmap to take back control of software and analytics.
Now, these are two areas which we did not want to give up to the vendors and be exposed to not only the lack of flexibility in using a vendor-driven model, but also the exposure to a very high cost structure because agility will translate to very high costs if you rely on vendors. Agility will translate to cost reduction and Operational Excellence if you rely on an internally built resource. T o achieve this, we have now built a 640-plus global pool of engineers with deep domain experience across a wide spectrum of software disciplines ranging from microservices, front-end engineering, RPA or robotic process automation, BSS experts, business support systems, that is IoT, UI/UX, QA, etc. T hese resources are located in three main markets in Sri Lanka where ADL or Axiata Digital Labs is headquartered and also in Malaysia and Indonesia, 100 resources approximately each in Malaysia and Indonesia and around 400 in Sri Lanka. Likewise, in analytics, like I mentioned, 270 resources, 153 of them who have now graduated through the training programs and 118 who are going through the process right at this moment, and that's across data science, data engineering, and data analysts. A new segment that we are now developing is the analytics translators. An analytics translator is someone who is well-versed both with the fundamentals of data science and analytics as well as with the business, and their core capability is to translate business problems into data science problems, and the hardcore data scientists would then take it to model building.
This enables the offshoring, if I were to use one of the capabilities that would come into the picture, but also to layer your capabilities in a way which there can be a production line kind of approach, including replication to analytics models. Finally, group-wide best practice has been built into this framework. I think I've spoken about most of these areas ranging from marketing and products, customers, customer experience, sales, enterprise, and then the simplification of processes inside the organization, IT modernization, data analytics, and now coming to the network. A lot of what I've spoken about so far has pointed to OPEX and Operational Excellence manifested in OPEX reduction.
Now, something I would like to end on is that the vistas ahead, provided you have the digital platforms ready and the digital capabilities of your people and also the architectures within the company, that you can begin to enjoy the CapEx advantages of new digital models. So two, I would like to highlight here in the IT domain, the fact that Axiata has broken up its IT architecture to expose APIs from legacy systems and thereby separate the legacy IT systems from a layer which comprises of microservices, high agility, high flexibility, zero CR, change request environments. Let me stop and explain that just for one moment. A change request is the order that you give an IT vendor each and every time you want to make a change, introduce a new product, add some new capability to your IT systems.
In a world where we are becoming increasingly software-driven, if you don't have a construct which enables you to do these things yourself, your cost as well as your time to market becomes a great challenge if your destiny lies in the hands of vendors purely because you have an external transaction going on every time you want to make even the smallest of changes. What Axiata has done in its BSS architecture is to leave the legacy systems, the hard plumbing as it is, expose some thousand-plus APIs which enable a new layer which was developed by Axiata called the Apigate DTE layer, which is an API-driven microservice layer enabling our telcos, our MNOs to develop their own microservices at speed and scale.
A microservice, again, that was developed in one OPCO can be transferred to any OPCO, creating a marketplace environment for a large number of telco microservices. T he CapEx reduction that has been already achieved here is very significant. G oing forward, we believe that this will be a springboard for optimization at scale as well as cost and CapEx excellence. Moving to the network layer, a lot of excitement around several global innovations in this area. I've just put the first three of these are quite interesting in order to achieve the higher performance broadband and 5G-ready network architectures. But there are many more. Taking the first, Dynamic Spectrum Sharing. Spectrum optimization or increasing your ROI or optimizing your ROI on spectrum investment is key.
Here, the newer technologies which enable spectrum to be used with dynamic flexibility, meaning that you can use the same band dynamically based on the traffic profile at that point in time between 2G, 3G, and 4G. Coming into the 5G era, this will become even more valuable. Our OpCos have begun over the last two years to use this technology extensively. This is where, as Tan Sri Jamal alluded to in his opening, we can go down to a segment, a much narrower segment in terms of our use of spectrum, just like we would do in terms of the service facilitation for a segment of customers. This would use our most valuable resource in the most optimum way. Core Network Function Virtualization or NFV, I think a term that you must be familiar with.
This is taking out the requirement for specialized hardware, separating the software of the core network elements from the hardware, enabling the use of the cloud for the installation of your network functions. F or example, your switching function could now be in the cloud on hardware that you don't even see. T his also increases flexibility, reduces costs. An emerging area is Open RAN, where even the base station now becomes software-driven. Y ou can buy your software for a base station function and install it on any type of hardware. And it can be open now end-to-end from your switching all the way to your base stations. We expect that the Open RAN constructs will really drive down the cost structure or the CapEx structure of mobile networks going forward. Globally, it's still at, I would say, not experimental, but advanced experimental stage.
Several MNOs have declared that they would be rolling out at least 10%-15% of their networks on Open RAN architectures. Now, this could be the great disruptor which changes the cost structure of our networks in terms of the CapEx. I n combination, if we look at attacking and disrupting the cost structure of the hardware, the software, opening up our microservices, making it flexible and very much like a digital system, then Tan Sri Jamal's vision that in three years, we will be very, very different. Not only in terms of how we do business, but how our infrastructure has been changed in flight while carrying millions of customers with us and delivering a service level which is adapted to a segment of one, quite a reality. So I'll stop there.