Thank you for standing by, and welcome to the Appen Limited FY 2019 half-year earnings conference call. All participants are in a listen-only mode. There will be a presentation followed by a question and answer session. If you do wish to ask a question, you will need to press the star key followed by the number 1 on your telephone keypad. I would now like to hand the conference over to Mr. Mark Brayan, CEO. Thank you. Please go ahead.
Thanks, Jodie, hello and welcome everybody to the conference call for our first half results for 2019. Thank you once again for your interest in the company and thanks for joining the call. I'm joined today by our CFO, Kevin Levine, you'll hear from him later on in the call. Per the intro, we'll walk through the presentation then we'll open the call for questions. Let's get underway. If you could turn to page four, first of all, for the results highlights. Look, the company's undoubtedly delivered once again for our customers, our staff, our crowd workers, and of course, our shareholders. Revenue for the half up to AUD 245.1 million. That's a 60% increase from the first half of last year. Underlying EBITDA of AUD 46.3 million was up 81%. At the same time, very pleasingly, our underlying EBITDA margins expanded from 16.8% to 18.9%.
This is as a result of economies of scale and our investments in tech-led productivity improvements as they took effect this half. Underlying NPAT was also up on the last half, or sorry, on the first half of last year, to be correct, by 67% to AUD 29.6 million. From a divisional perspective, these numbers are on page four relate to our core operating divisions and our independent of Figure Eight. Revenue for speech and image data of AUD 39.9 million was up 85% on the first half of last year, relevance data was up 48% to AUD 193.7 million. For those of you that have followed us for a while, speech and image data was formerly referred to as language resources, but we renamed it given its broader remit that now includes other data types such as image and video.
Relevance data refers, of course, to our former content relevance division. We are very pleased to declare a dividend of AUD 0.04 a share. This is partially franked. This is consistent with last year's interim dividend and in line with our ongoing review of capital management. Finally, Leapforce is now fully integrated and providing margin benefits, and Figure Eight is delivering on its thesis. More on both of these later in the presentation. Turning to page five and the divisional results. At the outset, just explaining that we have restructured the business to be closer to the customer. That is, we now sell speech and image data, which was sold from Language Resource, and content relevance data or relevance data, which was sold from the Content Relevance or CR Division. We share resources across these divisions now.
We've allocated those resources largely based on revenue, and we've restated the 2018 numbers on this same basis. The comparisons that you're getting on this page are on a like-for-like comparison. First, speech and image. This is really gathering speed. The first half of last year, we did AUD 21.5 million, and then we topped that in the second half of last year with a revenue result of AUD 29.9 million. That's now been bettered this half with revenue of AUD 39.9 million. In round figures, the business has moved to AUD 20 million to AUD 30 million to AUD 40 million in successive halves. I'm often asked, when will the world have enough speech data? I can confidently say, and this trend shows, that I don't think it will be for quite a while.
Our customers are continuing to ask for more data in many languages for new products and to improve their existing products. It's not just audio speech data. We also provide text data for language-based products such as chatbots, for example. The reason they need more data is that human speech and language is complex, and building technology that mimics it is a very hard problem to solve, and the solution lies in more data, and we're happy to play our role by continuing to provide very large volumes of high-quality data for all manner of speech and natural language product. Speech and Image, as the new name suggests, also provides image and video data. This is a growth area for us and one that requires special labeling technology. We're able to participate in it more fully now following the acquisition of Figure Eight.
Our relevance business continues to be the cornerstone of the company. Revenue of AUD 193.7 million in the first half was up 48% on the same period last year. This was largely from our anchor technology companies across new and existing projects, but we're also adding new relevance customers as well. Our technology-led productivity initiatives and economies of scale helped us grow our relevance margins. Remember that this is on a like-to-like comparison with the prior half or the prior corresponding half. We also continued at the same time to deliver high-quality work to our customers and fair wages to our proud workers. To page six now on Figure Eight. Our investment in Figure Eight is delivering on its thesis. It's accelerating and de-risking our technology strategy, and tech is essential to our business.
It's diversifying our revenue, and it's expanding our market, most notably into the government market, of which more later. Figure Eight technology and the engineering product and machine learning teams have been a very positive addition to the company. We are now firmly a data and machine learning technology company building on our established brand as a language company. The Figure Eight platform has proven to be deeper than originally thought, and its security features, including its SOC 2 compliance, and this is a certification that relates to the security, availability, integrity, and the data confidentiality of SaaS-based systems, is proving to be invaluable, particularly in the current climate of increasing scrutiny on data privacy. The achievement of planned synergies and the path to profitability are on track, as is our full year EBITDA outlook.
However, despite a strong first quarter, Q2 sales and retention were not where we would have liked them. The acquisition was distracting for staff, and it cost us some sales momentum. The sales team is also necessarily shifting to chase bigger deals, and some of these slipped outside the quarter. That said, we've implemented some remedies and added management support from Appen. Our new global head of sales, Jon Kondo, is spending much of his time with the Figure Eight team. Jon has a strong pedigree in sales growth and SaaS companies in particular, and his efforts are yielding benefits. We have over 30 cross-sell opportunities between Figure Eight and Appen, and the third quarter is off to a nice start.
We've revised our full year ARR target for Figure Eight. We note that the revision has an immaterial effect on the overall Appen results. Further, investors are protected by the earn-out structure. We view the current performance as temporary and retain a high conviction in Figure Eight and its thesis over the mid to long term. It's also worth noting that another company in our space called Scale recently raised capital at a valuation of $1 billion. That was well up from its previous valuation less than 18 months ago of $93 million. Regardless of what you think of the valuation, this shows genuine interest in the data annotation market and reinforces the value of Figure Eight to Appen into our future. We provide an updated customer cohort chart on page seven.
Chart shows our customers by annual cohort and the revenue for each cohort each successive year. We are clearly getting repeat and growing revenue from a number of these cohorts. The blue bars show speech and image data revenue for the first half of this year, almost on par with the full year for 2017. The red bars show relevance revenue. While results are dominated by 2 customer cohorts, the repeat nature of that revenue underscores the importance of the data we provide to our customers and the quality of the work and service that we deliver. You'll also notice that Figure Eight hits the chart for the first time, the yellow bar, for the first half of 2019. To page eight and an update on our technology strategy.
You'll recall that our investments in technology enable us to compete, deepen the moats around our business, and future-proof the business. I'm very pleased to say that our CTO Wilson Pang is delivering on that vision. The tech team incidentally includes engineers from both Appen and Figure Eight working side by side under Wilson's leadership. We rely on three core platforms illustrated on the top right-hand side of the page. Our crowd management platform, Appen Connect, acquired from Leapforce, helps us recruit, test, onboard, provide instruction, monitor quality, and pay our crowd workers. We're paying in the order of 40,000 workers per month, and that number's peaked at over 80,000 workers. This system is the foundation of a business of our scale and fundamental to our performance.
Our client workspace, acquired with Figure Eight, enables customers to fulfill their data needs through a SaaS-based interface. They can describe their data requirements and quality standards, send jobs to crowd workers, and manage data quality, all through an agile, web-based interface. Our annotation tools, some of which are from Figure Eight and some from Appen, are used by crowd workers to collect and annotate data to the client's requirements. We're investing into the usability and automation of these tools to improve the worker experience and of course, their productivity. We're delivering exciting new features across these platforms, including enhanced security features, very necessary in the current climate. We're also using advanced techniques such as machine learning to automate annotation tasks. We're already seeing productivity improvements from automated quality checking of transcription, for example.
Remember that our crowd is one of our greatest assets, but it's also a large expense. Improvements in productivity go to a better crowd experience and better margins. We need to invest in technology to run, improve, and future-proof the business, and the small table at the bottom of the page shows the substantial increase in investment from the first half of last year. This is an important part of our transition to a tech-led business. Over to page nine and a brief update on the acquisitions. Leapforce is now fully integrated into Appen. The Appen Connect system came with Leapforce, and in addition to the system's recruiting and crowd management capabilities, we gained comprehensive project and resource management functionality as well as quality management and reporting features. The integration of Leapforce included a root and branch review of all of our crowd and customer-facing processes.
The implementation of Appen Connect to codify these processes, and then comprehensive training and change management to enable our recruiting and customer teams. Appen Connect also enabled the restructure around the customer where we share resources now across speech and image projects, relevance projects, et cetera. It was a non-trivial exercise, but we're already seeing the resulting productivity benefits and have set the business up for far greater scale. We've yet to fully embark on the integration of Figure Eight. It's running as is under the leadership of its COO, Becky Scott, to deliver the earn-out. Appen management is leaning in to coach, scope the integration, and set the business up for future success. Independent of the integration, Figure Eight is providing us with substantial benefits, including giving us a great position in the government market. This is highlighted on page 10.
The Figure Eight platform has been deployed behind government firewalls for large-scale data annotation for government AI projects. This is no mean feat. It required rigorous testing to the highest technical security standards and swift and agile responses from Figure Eight engineers to respond to issues and satisfy these standards. It's interesting that the government came to Silicon Valley and Figure Eight rather than traditional government tech providers, and shows both the uniqueness of Figure Eight's solution and the government's intent to invest in AI. Also, the secure deployment of Figure Eight is a high barrier to entry, and it sets us up to fully serve the customers and participate in what could be a substantial opportunity for the company. Page 11, we continue to invest in China, the second-largest AI market, and may be the largest one day.
We welcome Dr. Roc Tian to the management team as our head of China. Roc has a strong technology pedigree through his career with IBM and HP in China, as well as a PhD in computer science. His background with U.S. firms in China brings valuable cross-cultural business experience, and he operates with a high degree of communication. He's also hit the ground running, building the team, and working with our customers. Our data labeling facility in Wuxi, which is west of Shanghai, is implemented and operable. It's similar to our Philippine center, but currently at a smaller scale. I'd now like to hand it over to Kevin to take us through the financial pages of the presentation.
Thanks, Mark. Once again, it is a real privilege to report on such great results. The revenue increase of 60% has been driven by both divisions as well as Figure Eight. Out of the 60% growth, organic growth represented 53% and Figure Eight represented 7%. Growth of 85% in speech and image over the prior period was a standout and mainly came from expansion of existing projects continuing on from the previous half. There was also some growth in new projects. Relevance growth, excluding Figure Eight of 48%, was driven by major customers with growth in existing and new projects. Underlying EBITDA was up 81%. This was achieved by efficiencies from the Leapforce integration, as well as efficiencies from other process improvement initiatives. As we get bigger, we get better, and we continue to derive economies of scale. Underlying NPAT was up 67%.
This was mainly impacted by movement in effective tax rates to 28.2%, from 21.2% in the prior comparative period, where the prior period rate was abnormally low. In the prior period, the share-based payment deduction was significantly larger than the share-based payment expense, that gets added back to taxable income. That resulted in an abnormally low tax expense. Moving on to slide 14 and to the balance sheet. Appen continues to strengthen its balance sheet. Receivables have increased due to increased volumes as well as the addition of Figure Eight. Non-current assets have increased significantly, mainly as a result of the Figure Eight acquisition. Of this increase of AUD 311 million, goodwill associated with the Figure Eight acquisition represents AUD 281 million, and capitalization of rental leases represents AUD 22 million.
Based on the revised Figure Eight's ARR forecast, the corresponding estimated future value earn-out payment rate is AUD 36 million-AUD 50 million. The estimated earn-out liability has been booked at the low end of this range in order to be conservative. Debt has significantly reduced. AUD 45 million of debt was repaid during the period. AUD 21 million was repaid as part of the capital raise in March, and a balance of AUD 24 million was sourced from our own cash reserves. As a result of the low debt levels, Appen is in a net debt negative position of AUD 59 million. Essentially, we have more cash than debt. This compares with net debt leverage of 1.1 times at the time of the Leapforce acquisition in December 2017 and 0.26 times at December 2018. Moving on to slide 15, our cash flow.
Appen continues to execute well on converting earnings to cash and thereby increasing cash balance. Cash increased by AUD 40 million in the period, driven by a 98% increase in cash flow from operation. Cash was directed to repay debt, fund CapEx, and pay dividends in the period. Cash flow conversion increased to 92% from 84%. On slide 16, we talk about the impact of currency and the Appen performance benefited from a weak Australian dollar in the period. Excluding the Forex translation tailwinds, revenue and underlying EBITDA increased by 47% and 63% respectively. The EBITDA uplift that came from the Forex translation, the spread between that and Appen would be in accordance with each division's relevant earnings. I'm going to hand you back to Mark. Thank you.
Thanks, Kevin. It's page 16 now. The AI market continues to grow very strongly. There's no doubt of that. There are many indicators, and some are on the page. Gartner, for example, a leading tech analyst that many of you know, estimates that the number of AI projects per organization will grow at over 100% a year. KPMG found that most businesses have some adoption plans for AI, varying in maturity from pilot projects, to scaling up, to in production and industrialized. This is across a wide variety of use cases. Now, none of this growth is possible without high-quality training data, and we're playing our role admirably in this regard. Over to page 17. We're uniquely and strongly positioned to win in this market. We have unparalleled suite of technologies, a proven and highly scalable crowd capability, and an enviable track record.
As previously stated, our speech and image growth rate of, in round numbers, AUD 20 million to AUD 30 million to AUD 40 million over the last three halves illustrates this demand for data and our ability to deliver at scale. Our customers are demanding, but we meet those demands, and we continually improve to exceed them and meet their needs. Finally, to page 18 for the summary and the outlook. We're really pleased to present the results and assure you that we remain committed to strengthening our position and continuing the company's growth in the dynamic market for AI training data. We'll continue to invest in technology, sales, and marketing in the government markets in China in particular, to achieve these objectives. Our 2019 order book is in good shape, with year-to-date revenue plus orders in hand for delivery in 2019 at approximately AUD 380 million at mid-August. This includes Figure Eight.
Our full-year underlying EBITDA outlook for 2019 is trending to the upper end of the range of AUD 85 million-AUD 90 million. This also includes Figure Eight. A few notes on the earnings figures. The second half is struck at a U.S. exchange rate of $0.74. We realize this is higher than the rate today, but it's our practice to maintain the rate that we struck at the beginning of the year for consistency. Note also that we took three months of Figure Eight losses in the first half, amounting to AUD 2.6 million, and we'll take six months of losses in the second half. The first half result benefited from a Forex gain of AUD 4.7 million.
Before we open for questions, I would like to thank our customers for their support, our global crowd for the work they do for us, and the hard work, commitment, and friendship of everyone at Appen and Figure Eight. With a special thank you to Kevin for everything that he does for everyone on the call and for all of our shareholders. Pleasure working with such a great team, and a pleasure to report the results of their efforts to all of you. Thank you. I'll now hand the call back to Jodie to open it up for questions.
Thank you very much. If you do wish to ask a question, please press the star key then one on your telephone and wait for your name to be announced. If you wish to cancel your request, please press the star key then two. If you are on a speakerphone, please pick up the handset to ask your question. Thank you. Your first question is from Quinn Pierson from Credit Suisse. Go ahead, thank you.
Hi. Good morning. I guess just firstly on Figure Eight and the slowing of the revenue growth there. You've highlighted some transaction disruption. You've highlighted some, I guess, chasing some of the larger deals, presumably, that would've been kind of known or understood when you were providing original guidance on that ARR growth rate. Could you maybe just talk us through kind of what is new or what has changed since March that has slowed the revenue growth momentum down?
Yeah. Hi, Quinn. I think as we said, the transaction itself was more destabilizing than we thought for the team. It's a youngish team in a very dynamic environment in the Bay Area, and a lot of people were sort of fearful for their jobs initially, and that kind of put them into a little bit of a funk, unfortunately. The good news was that the staff retention rates were high, so it didn't result in us losing staff, but it did result in some deal disruption. The other thing, and you may know this also, is that these businesses are very highly tuned towards the end of the quarter. What we see in March and in April, and even May, can be very different to what we see come the 30th of June in this instance.
The end of the quarter is when the large deals are signed or resigned. In this case, some of them not signed. There was a bit of a swift transition to the visibility. The good news is, going into Q3, we're off to a fast start. Retention rates are high. Jon Kondo, as I mentioned, has settled the team down and starting to add a lot of value there. We're pretty confident in things going forward. We also are supportive of their push to focus on larger deals because we think the larger deals are out there, and we think that's how they're going to add a lot more value to the business.
That's helpful. Secondly, and related to that, I think I heard you say that you're still confident on the previous targets, which was second half 2020 profitability for Figure Eight standalone. I guess could you just kind of confirm I heard that correctly? If that is the case, presumably that would mean your FY 2020 revenue growth would be even higher than you had previously thought? In other words, kind of make up for the FY 2019 disruption. Is that how to be thinking about it?
Definitely you're correct on the first progression on profit. In terms of the full year next year, as we go into our planning towards the end of this year and as we bring the businesses together, clearly we'll get an accelerating effect from the integration into Appen. Exactly how that will map itself is a little early to say at this point.
Just to be clear, if you're still confident in the second half 2020 profitability, is there a chance that maybe the cost base would be lower to get there? Not necessarily the revenue hitting original expectations, but kind of the net profit. Is that part of it?
Yeah. It's a combination of revenue and cost management. We're managing the cost a little more prudently, perhaps, than the business did on its own. Obviously we're confident in the Q3 and the Q4 revenue to get us to that EBITDA target by the end of the year.
Wonderful, thanks. Lastly, on your full year guidance, you're tracking towards the top end of the range at the AUD 90 million. If we look at AUD 46 million of first half EBITDA, that implies about AUD 44 million of second half EBITDA. I'm cognizant there is a little bit of annualization of Figure Eight losses, but even including that, call it a somewhat flattish outcome there. Are there any kind of first half projects that would be rolling off to be cognizant of? Or is it more, I guess, a matter of conservatism where you're not making, I guess, potential or likely new work later in the year? I'm just trying to reconcile the first half, second half profit split.
Yeah. You've got to take the impact of the foreign exchange into account, Quinn. We record actual results at the exchange rate on the day, but we forecast the forward period, in this case, the second half, at $0.74. There's a dampening effect on the second half forecast due to currency. As I explained or stated in the presentation, that's just been our practice. This year, it's happened to swing quite a bit. There was a $4.7 million benefit from currency in the first half. You can assume that there'll be some currency benefit in the second half. Overall, when you take into account currency benefits and the Figure Eight losses, we see growth in the second half on the first half.
Yeah. Just at spot right now for our Forex, what would be the implied earnings uplift, if you were to use that?
I think you're gonna have to do your own sums, Quinn.
Will do. Thank you.
Quinn, what you can do is you can look at the movement in H1, and you can see how that correlated to the 4.7. Essentially, if the currency movement is around that or similar, well, then that could point towards a similar kind of outcome.
Helpful. Thank you. [inaudible].
Thanks.
Thank you. Your next question is from Garry Sher from Royal Bank of Canada. Go ahead, thank you.
Hi, Mark and Kevin. Just a question around F8, particularly around the earn-out. It sounds like that earn-out has dropped materially. Can you just remind us what it was previously? I think you're now guiding to AUD 36 million-AUD 50 million earn-out.
Garry. Basically, the earn-out is based on the ARR contracted to the end of December. Essentially what this means is any deals that really get signed up before the end of the year go into the mix. The guidance we provided previously was $62 million.
Sorry, just to clarify USD 60 million-USD 80 million. You've got a number of figures here around Figure Eight, in relation to ARR, and you've talked about being below plan at a certain line and then at plan at EBITDA line. I wonder just for clarity, are you able to just tell us what those plans were? You're now saying ARR has dropped, so I'm just wondering what you were going for before. Also, you're now saying that EBITDA remains in a plan range. Can you just maybe be a bit clearer on what those figures are?
Essentially, yeah. When the word plan, you should read into the guidance we provided at the time when we announced the acquisition. Basically, we gave 2018 numbers, for Figure Eight, and then we basically said, okay, from an EBITDA perspective, what we expected is it was gonna be a 25%-35% reduction in the 2018 number. We're confirming that position. There's no change in that. Once again, the note when we're talking about the ARR, it references back to the guidance that we gave at the time of the acquisition.
All right. Your year-to-date revenue plus orders on hand, lastly, that ratio, in terms of delivery to revenue, was about 70%. Should we be assuming a similar ratio conversion for 2019?
There's a few factors that impacted, Garry. As we win more customers, we get a better spread of revenue through the year, which is to say that there are bigger orders to come year on year. It's a good starting point. It's going to vary year on year, but it's a good starting point.
Final question on capital management. You've stated an ongoing review of capital management priorities, including a dividend policy. You've just got net cash of AUD 59 million. Can you just remind us what your current dividend policy is, and maybe give us a sense of whether you're planning to build up more of a cash war chest for future growth? Or are we thinking that's not the case, and maybe you're going to lift the payout ratio? I just wanted to be a bit clearer around that, if possible.
Yeah. We don't have a stated policy on dividend, Garry. We've got some debt now, so we need to manage that. To the extent to which we can find attractive opportunity for acquisition, having more cash for that is always handy, and also having cash to invest into such markets as the U.S. government market and the Chinese market. I think you certainly get the view from us that we like to do things reasonably conservatively, and we think that using that cash for investment is probably a good thing to do.
Perfect. Thank you.
Thank you.
Thank you. Your next question is from Paul Mason of Evans and Partners. Go ahead, thank you.
Hey, guys. Just a few from me. The first one, your investment in engineering and CapEx at AUD 13.1 million. Can I just get some context around that? I'd thought you'd sort of guided to about an AUD 6 million annual budget for that for this year. I know I've got an AUD 6 million number from one of your presentations, but maybe just explain the differences going on there. Is that just actually much more expense, or is that AUD 6 million like an incremental thing and I'm officially interpreting history wrong?
Hi, Paul. The AUD 6 million figure was initially savings from the integration of Leapforce that were being plowed into R&D prior to the acquisition of Figure Eight. When we made the acquisition of Figure Eight, we were still using most of that AUD 6 million for R&D. Of course, we ended up, or we are now spending more than the AUD 6 million on R&D because we have a larger team of engineers and data scientists and product managers. That AUD 13 million figure refers to the total spend in engineering across Appen and Figure Eight. The six is buried within it. The reason it's gone up so dramatically is because we've added the Figure Eight engineering team.
Okay, great. Just on margins in speech and image. It was at the second half last year, you guys saw effectively a very down year. I think your explanation at the time was that a whole bunch of government work or secure facility work, at least anyway, had temporarily disappeared, and you were flagging it was coming back. Is that the key driver of the margin recovery in this half, or is that more attributable to the speech and image recognition stuff, the AI that's gone in, as opposed to the more legacy half of that business that were lower margins historically?
Yeah. We haven't seen a strong return to that previous very high-margin government work. The revenue growth in speech and image is coming largely from the technology sector, from commercial clients. The margins are generally consistent. Just for clarity, the government work that I referred to with Figure Eight is in a different part of the government altogether. It's not that very high-value speech or very high-margin speech work that we have historically enjoyed. Overall, the revenue growth coming from the tech sector and the margins are about where they were.
Okay. Just the last one from me. On the Leapforce integration, just noting what you'd said in prior answers that you'd flagged AUD 6 million of savings for this year. Just going through the slide seven, it sort of indicates that in relevance, about 50% is done at the moment. In speech, image, and video, you're expecting stuff to migrate in Q2. Just two elements to this, if you could. Just in terms of where you said Q1 and Q2, is that a reference to financial year Q1 and Q2 or calendar year Q1 and Q2? Like, Q1 ending in September or does it end in March? Secondarily, given that there is clearly it's not 100% run rate, could you give us an idea of the annualized full efficiency benefits, given AUD 6 million has clearly got a ramp profile for it?
Sorry, Paul. I don't understand your question. Can you break it down a bit? There's a whole bunch of stuff in there.
Maybe I'll ask it as two questions then. Just the first one, on slide seven, you've made comments about relevance projects getting onto Appen Connect 50% now or by the end of Q1, and then for speech, image, video work, end of Q2 on Appen Connect. Can you just clarify what Q1 and Q2 means in terms of the exact dates how you're topping the year off?
I think you might be using an older presentation, Paul.
Okay. All right. Never mind. All right. Maybe I'll send you an email afterwards, but I'll just go back and check some notes on this.
Okay.
All right. All right. That's all from me for now. Thanks.
Thank you.
Thank you. Your next question is from Josh Kannourakis from UBS. Go ahead, thank you.
Hi, Mark and Kevin.
Hey, Josh.
Just first question, around the relevance division. Obviously you mentioned some restatement there, but the delta in margins year-on-year are actually really large. I guess, just trying to understand a little bit more context around how those margins changed from second half 2018 as well, and whether you expect sort of a continued benefit, and just maybe a little bit around where that benefit's coming from. Obviously, a strong result.
Yeah, thanks. There's three sources of the benefit. One is a step change, the other two should continue to yield benefits going forward. The first is just the full integration of the Leapforce work. Leapforce operated at higher margins than Appen historically due to its high degree of efficiency. Just blending the Leapforce work and the Appen work gave us a lift of margins, the first source. The second source is technology-led productivity improvements. We've implemented the Leapforce platform now across all of our projects, our speech and image and our relevance projects. That just improves efficiency and goes to margin. The third source is just economies of scale. The bigger the projects get, the bigger the business gets. We get to share cost, and that goes to margin as well.
Got it. In terms of the continued benefit throughout the year, have we seen a full hit of, I guess, especially the tech-led and obviously Leapforce-led benefit within this first half? Is there any more benefit to come in the second half?
There should be, to the extent that we don't get price pressure. Remember that the thesis around technology is to improve productivity through technology and position us to manage price pressure if it should come along. We've always targeted a group level margins in the mid-teens. We've exceeded that at the moment, and we'll continue to deliver technology that improves productivity. Of course, we can't predict price pressure at this point.
Got it. That was actually my next question. Just to get a little bit of context, Mark, around the customer environment and what you're seeing out there in terms of pricing across some of your large clients.
The larger the clients, the greater the demand for volume discount, and we work with our clients in that regard. Most of the pressure that we see comes from demand more so than price at this point. We are being proactive with our customers to provide them with good pricing and they're very happy in that regard. We also are seeing most of our pressure on the demand side and our ability to scale up with crowd workers, for example, is very helpful in that regard.
Got it. Just following on from that, though, in terms of, I guess, the competitive landscape, are you seeing any new entrants or in your core markets? Obviously, we talked about some other guys on the autonomous side like Scale AI and the like, which you mentioned earlier. In your sort of core relevance markets in social media and search, it's still the same players, no new entrants at the moment?
That's correct.
Okay, great. Just final one, just in terms of the speech and image division, are you able to give any context around just what amount of sort of speech versus image? I imagine image is small, but just to try and get some context on how that part of the business has been going.
Yeah, the bulk of it is speech work. As I said in the presentation, our ability to participate at scale in the image and video world requires certain technologies which we now have with the acquisition of Figure Eight. Speech and image is predominantly speech and other natural language datas. We're confident that we'll see more image and video data going forward.
Got it. Just in terms of the cost base corporate overhead, is it fair to expect that to sort of continue at a similar run rate into the second half? Is there any other sort of movements we should be aware about?
Yeah, Josh, I think that that's indicative of normalization. Obviously, if there's special areas we need to invest in to further growth, we'll do that. If you look at, we finished full year 2018 with AUD 11.1 million, AUD six and a half for the first half. That's indicative of the best view of what we have right now of corporate overhead.
Yeah. Definitely final one from me. I think Paul mentioned the AUD 13 million invested in the business. Can you just give a bit of context around what was, I guess, capitalized versus what was expense of that? Was that just a pure expense number?
That's the total investment.
Yeah.
If you go back to the guidance in general policy, it's a three-year amortization policy. Essentially work on essentially one-third of that is being capitalized and two-thirds is being expensed.
Perfect. Thanks, guys.
Yeah. Thanks.
Thank you. Your next question is from Adam Dellaverde from Taylor Collison. Go ahead, thank you.
Hi, guys. Thanks for taking my questions. Just as we talk about retaining cash for acquisitions, I'm just interested in your thoughts in terms of deploying capital into China, maybe in the context of one of the big tech companies in the U.S. being criticized for aiding the Chinese government indirectly.
Hi, Adam. Yeah, I don't know how to answer that question. We're going to be fairly prudent in the way we deploy cash into China to manage the risk of our business. Yeah, I'm not 100% sure what you're after, but I will say that we're going to be fairly prudent in the way we deploy the company's cash in China.
Sorry. I guess I'm just alert to, on the risk side, there's this fantastic business growing like crazy with sort of, let's say, Western customer base. You push into China, how do you protect systems? How do you protect data? How do you vet to make sure that those customers are happy, and all of the work you're doing for them is sort of siloed? How do you hire, I guess?
Yeah. No, that makes sense. We are very aware of the risk of the nature of our business in China, being a data-based business. We have deployed a completely separate tech stack into China. Our core platforms, our Appen Connect platform and our Figure Eight platform, are not being sold into China or being used in China. China has its own tech stack, and so that keeps the IP and the data completely separate. We're also setting up China as pretty much a standalone entity for reasons of protection of IP and data. Also to be able to respond to the local culture and the pace of business there. We've been very open and transparent with our customers about our activities in China, and they're very pleased with the way that we're structuring things.
They maintain some interest in Chinese data, so they're pleased that we're there, but they're also doubly pleased that we're going to the trouble to structure things in such an air-gapped fashion.
Great. I guess in terms of the established customer base, which has shown amazing organic growth again, are you sensing any change in the way they are tendering, I guess in some of them you're exclusive? Are you hanging on to that exclusivity to the best of your knowledge?
Again, I don't think we've ever said that we have exclusive relationships. I will say that, and just echoing my earlier comment around demand, the most important thing for the customers is a reliable supply of high-quality data at speed for a good price. So long as we continue to deliver, our customers continue to be very happy. Per the earlier question, I think from Josh, or maybe Paul, there was a question about competition. We don't see any change in the competitive landscape at the large scale. There are smaller startups on a regular basis providing annotation services, but not to the level or the scale or the quality that our large customers thus far are comfortable trying out.
Great. I guess just as you retain some capital and look to deploy it into acquisitions, if I sort of think about the acquisitions you've done so far, there's been an element of, I guess, productivity, technology benefit to the core business and then something new on the other end, maybe a new customer or maybe a new SaaS business. As you're looking forward, do you think it'll be a similar style where you kind of justify part of it based on improvement to the core? Or is there some new capability or some new SaaS or new sector that you're looking at when you talk about that kind of opportunity?
Yeah. I think we've got a pretty good set of technologies at the moment. At the moment, I think we're pretty solid there. We'd love to find the opportunity to add customers or add markets or add a related capability. I don't think we'll be going for a purely tech play at this point. It's just my sense. We keep our eyes open for opportunities. I think we've got a pretty good set of technologies at the moment.
Great. Thank you.
Thank you. Your next question is from Stella Wang, private investor. Go ahead, thank you.
Good morning, guys. Great choice of Wuxi as your processing center. Great location, close to Shanghai and Hangzhou and not as expensive. I got two questions. The first one is on R&D, AUD 30 million. I think the previously guided equivalent or comparable is about AUD 12 million. Should we expect the AUD 30 million to continue, or is there any synergy you can strip out of that? Could you give us a bit of an update guidance on that front?
Yeah. I'm not sure of the AUD 12 million number, Stella, but certainly, R&D is important for the business, and I think we'll see R&D being the future of the business going forward. It's not going to jump up at the rate that it has as that table illustrates, because we've started from a pretty low base and also with the acquisition of Figure Eight, that has a lot of tax costs. We will continue to invest in R&D going forward. It's necessary for the future growth of the business.
Okay. Very nice. Staying in the last question, in your relevant segment, could you comment on how much of that work is done on data sourced by your customers from their end users? As we all know, there are lots of regulatory interrogation into transparency and privacy issues there. Just want to get some idea of whether that would impact the company in any shape or form in a short or longer term.
Yeah, you're correct. Data privacy is very important and we always have and will continue to comply with the privacy regulations, and we work closely with our customers to ensure that them and us comply with the privacy regulations. From time to time, there are projects that the process is adjusted to comply, but we don't see a material difference in the work we're doing, the projects we're working on, and the trajectory of the business. It is something we're super aware of.
Okay. We shouldn't expect any material changes in that area for you?
We don't see it, no.
No. Great. Thank you. That's all for me.
Thanks, Stella.
Thank you. Once again, if you do wish to ask a question, please press the star key and one on your telephone and wait for your name to be announced. There are no further questions at this time. I will now hand back to Mark Brayan for any closing remarks.
Thanks, Jodie. Thanks again to everybody for joining the call and for your interest in the company. Just to reinforce that we're very pleased to bring you these results. We're very positive about not only these results, but the trajectory of the business, and looking forward to meeting many of you face-to-face over the coming roadshow. Thanks once again.
Thank you very much. That does conclude our conference for today. Thank you all for participating. You may now disconnect your lines.