Thank you for standing by. Welcome to the Appen Limited FY 2020 full-year results conference call. All participants are in a listen-only mode. Today's call will be one hour in duration. There will be a presentation followed by a question and answer session. If you wish to ask a question, you will need to press the star key by the number one on your telephone keypad. Please register your question as early as possible. I would now like to hand the conference over to Mr. Mark Brayan, CEO. Please go ahead.
Yeah. Thank you very much. Hello, everybody. Welcome to the conference call for Appen's results for the full year ending December 31st, 2020. My name is Mark Brayan, I'm the Chief Executive. I'm joined today by our Chief Financial Officer, Kevin Levine, and Head of Investor Relations, Linda Carroll. Our results presentation was loaded this morning. It is available on the ASX website. I'll be referring to that throughout. We will take questions after the presentation. To page three to commence. Appen makes artificial intelligence work in the real world. Products that use AI are developed using a process called machine learning, in which an algorithm learns patterns that enable it to mimic human functions, such as reading, speech, vision, and making choices. The examples are known as training data. The more data, the better the AI performs.
Importantly, the training data must match the use case and the real-world situation. The AI won't function correctly if the data doesn't represent the real world. Appen is the world's largest provider of AI training data. We leverage our unique technology and our crowd of over 1 million global workers to collect and prepare large volumes of high-quality training data for the world's leading technology, automotive, financial services, retail, and healthcare companies, as well as government agencies. We do this across a variety of use cases, some of which are on page four. We provide human relevance data to search engine providers to ensure they serve up relevant information and advertisements and do so consistently and, importantly, without bias. We provide two-dimensional and three-dimensional image training data to autonomous vehicle companies for their self-driving car initiatives, as well as speech data so drivers can safely access technology hands-free.
We're providing image data for augmented and virtual reality that will train the systems to recognize the actions of the users. This is an exciting new area for us that opens up markets such as gaming. We have a long track record and considerable expertise in speech and natural language data, and we continue to support the development of chatbots and other customer experience technologies. Finally, e-commerce, a growth area in the recent pandemic, requires our expertise in search and natural language processing. If you could turn to page five for our 2020 full-year results. We're very pleased to deliver another solid year of growth for shareholders. The growth was not to our usual standards for a reason we'll get to, but we are nonetheless very pleased to deliver this result given all the challenges of 2020. To borrow a sporting metaphor, 2020 was a year of two halves.
We had a very strong first half with revenue up 25% on the back of relevance revenue growth and an FX tailwind. The second half was a different story as we know from the December trading update. Revenue was slightly down on the first half due to factors that I'll go into. We also faced a stiff FX headwind. Overall, though, 2020 growth was solid on 2019. Our long-term prospects remain very strong. Revenue in 2020 was up 12% to AUD 599.9 million. Underlying EBITDA of AUD 108.6 million was up 8% at a margin of 18.1%. We're pleased to announce a full-year dividend of AUD 0.10 per share, up 11% on the 2019 dividend.
In addition to the FX headwind, and you can see the effect of that on slide 17, the second half was impacted by a number of pandemic-related factors, including COVID-19 resulted in an unsurprising slowdown in online advertising in the middle of the year. Our major customers rely on advertising as their major source of revenue, and hence they reacted by deferring some projects and reallocating resources to new product developments to lessen their reliance on advertising. This, in turn, impacted our revenue to the extent that we did not see the uplift in revenue from our major customers in Q4 that we have seen in other years. Further, our B2B selling motion was impacted by the shift to working from home caused by the pandemic.
Our sales teams and our customers took a quarter or two to move from survival mode, including setting up at-home offices and technology, to business as close as possible to usual mode and a return to pre-pandemic deal cadence. Finally, many of our small customers were and continue to be impacted by the pandemic, although the overall impact of them on us is not material in the scheme of things. If you could turn to page six. 2020 was a breakout year for new customer wins. We added 136 new customers last year, and you can clearly see the slowdown caused by the pandemic and the strong uplift in Q4 as the sales motion recovered. These wins were across all data modalities and many use cases, and we'll share some of them later in the presentation.
These wins were enabled by our investments in sales and marketing, as well as our annotation platform, which we acquired with Figure Eight in 2019. These customers are small and/or early in their AI journey, and hence they won't have a material impact on our revenue in the near term, but they provide a foundation to build upon and give us confidence that our offerings are attractive and competitive in the market. We also had a substantial increase in new projects across our top five customers. Many of these were enabled by the customer's use of our annotation platform. They can cover more use cases and more data types using our platform than their own technology, and the platform integrates with their operation, improving productivity and quality, as well as increasing customer retention.
These new projects are doubly important because these are the projects that our customers are investing in to reduce their reliance on advertising. That is, the projects where resources were reallocated to late last year. These products are early in their life cycle and haven't replaced the revenue impact we saw in the second half, but they will grow in time. We're also very pleased to report high growth in China. Revenue is up 60% quarter on quarter, and we count China's major technology companies as our customers, as well as many others in autonomous vehicles, health, and education. To page seven, and we're also pleased to report a material increase in committed revenue to 31% of our total revenue in the second half, up from 12% in the first. The increase in committed revenue is significant because it gives us more revenue visibility and predictability over time.
The chart on the left shows annual contract value or ACV at the end of each half, as well as at the 1st of February. Clearly, we had some substantial renewals early in 2021, but the dip from the first half to the second half 2020 shows the impact of the pandemic on our smaller customers, many of whom have committed contracts with us. While we continue to grow our committed revenue, our cohort chart on page eight shows a high degree of revenue repeatability year-on-year, regardless of the contract type. This is because our customers rely on us for an ongoing supply of training data to ensure their products stay relevant and improve in quality and utility. Now, clearly, they have some flexibility in their data needs as we experienced in the second half, but the long-term trend gives us confidence.
In regards to our major customers and projects that were deferred or impacted by resource reallocation, we are seeing most of the material projects that were deferred in the second half recommence in the first half of 2021. Present indications suggest a steady return of the projects that were impacted late in 2021. On page nine now, relevance continues to be the bedrock of the business. Revenue was up 15%, EBITDA up 8%, with the second-half impact that is already explained. I should also point out at this stage that our work-from-home delivery model remained resilient through the pandemic, ensuring that we are always able to support our customer requirements. The chart on the right shows healthy year-on-year growth to relevance. This is due to a few factors. Firstly, AI that relies on relevance data requires constant refresh to stay relevant to users and unbiased.
Secondly, delivering relevance data at scale requires a highly specialized set of capabilities that we've developed over many years. These include our crowd management technology, our annotation platform, and our multinational multilingual crowd of over 1 million people in over 170 countries. The crowd is especially important. Our customers require data that is culturally and linguistically accurate and only available from in-country workers. It's not amenable to delivery from a single low-cost location, nor is it practical or cost-effective to in-source relevance work, especially when you consider that we paid relevance workers in 87 countries last year. The crowd also provides the necessary human aspect to relevance data, which is essentially a human's judgment between choices, making it near impossible to automate. Our relevance customers are some of the smartest data scientists on the planet, and they would have automated it already if they could have.
We can improve productivity with our crowd platform, however. We're using AI, for example, to automatically allocate tasks to workers based on their skills and track record. It's not unlike a product recommendation that you get in an online store. More on this later. The specialized nature of delivering relevance data at scale means there are just two meaningful players in this space, us and one other. We don't see change to the competitive environment, especially in relevance, and the general hype around our space is amongst companies we know well and have known for some time. Page 10 in speech and image, which was down on a breakout year last year, but still on an upward trend, clearly visible in the chart on the right-hand side of the page. AI products that mimic the speech and image capabilities of humans share some data requirements with relevance.
The data must fit the use case, represent the world, and be of high quality. The refresh rate is lower. An AI product that, for example, automatically recognizes speech in a particular language, requires modest amounts of data to account for changes like new words or accents or acoustic conditions, and only needs a lot of data when it needs to support a new language. As such, data needs are more cyclical and tied to the product's development life cycle. Hence, we've seen some ups and downs in the almost 25 years that we've been providing training data, but the long-term trend has always been positive. It's worth noting also that we are doing more image and video work, including in augmented and virtual reality. This is very exciting and could be a growth area with applications in many markets, including gaming.
On page 11, we outline our growth investments for 2020, which were dominated by sales and marketing, most of which was in the first half as we ramped up our go-to-market capability. The chart also shows a tidy FX gain from restating U.S.-denominated debt and our hedge book. On page 12, in our China business, we're pleased to report rapid revenue growth in China at 60% quarter-on-quarter. Our gross margins are improving. This progress validates our market thesis and strategy. We count China's major technology companies as customers across multiple projects. We've won other customers in the autonomous vehicle, health, and education technology sectors. We're working in all data modalities, speech, relevance, image, video, and lidar. It's also very pleasing to win speech work in local dialects against local competitors.
On page 13, we highlight some sales and marketing successes, including 136 new customers, and we increased the number of projects in our top five customers by 34%. Both outcomes are significant. The high number of new customers, although small and/or early stage, validate the attractiveness and competitiveness of our products and services and provide a solid foundation for future growth. The new product wins in our major customers are the very projects that drew resources from our major programs late in the year. While early stage, they are in exciting new areas and some could be substantial. They're also driving more volume through our annotation platform, validating the value that our customers derive from it and strengthening our relationship with them. The use cases on the right are rich and varied and all new from the ones presented at the half.
Along with more speech and natural language customers in automotive, air traffic control, and financial services, we're doing more image work in AR, VR, autonomous vehicles, and as well as a lot of document-based OCR, optical character recognition work, for applications that extract information from scanned documents, such as invoices or expense receipts. Page 14 includes some technology highlights. There are a few things here that I'm very excited about. Our crowd management platform, Appen Connect, is maturing in its security, scalability, and feature set. We now have an AI engine that automatically matches workers with tasks and greatly accelerates project ramp-up time to value. We're also using AI to catch fraudulent and mischievous workers, which saves us money and improves data quality, which is both good for our customer and for us. The resulting efficiencies improve our productivity and will, over time, improve margins.
Our annotation platform is also benefiting from AI. We have AI-assisted annotation or prelabeling for multiple data modalities such as text, speech, image, video, and lidar. They increase the speed of annotation by up to six times. We're also using AI for some labor-intensive data prep, such as splitting larger datasets into the discrete data points that our customers require. This lowers our unit costs, yielding higher margins and enabling more competitive pricing should we need it. We've recently launched a mobile app enabling crowd workers to engage with us on their phones and tablets, which greatly improves their utility and experience. Our government team on page 15 is doing well, having faced multiple challenges in 2020, including the pandemic, the U.S. presidential election, and Brexit. Growth was a little slower than hoped in 2020. We're optimistic for 2021. The government market is still fundamentally attractive.
I'd now like to hand it over to Kevin to talk you through the financial slides.
Thank you, Mark, and hello to everyone. Total revenue is up 12% on the prior corresponding period, driven by continued strong growth in relevance, which is up 15%. Relevance benefited from increased demand for data annotation in both existing and new projects with existing customers. Our normal historical revenue growth pattern, which sees us skew to the end of the year, was impacted by the strong AUD in the second half, as well as our major customers' response to COVID-19 and the changes to their activities and priorities. Overall, though, our major customers have been solid and a source of strength during the pandemic. Speech and image revenue is down 10%. Speech and image products are cyclical in nature, heavily dependent on customer timing, investment and product life cycles, and require less ongoing data refresh than relevance projects.
As a result, this can significantly influence performance on a half-on-half, year-on-year basis. This was evidenced by some significant project completions and negotiations in FY 2019, awaiting the next product and investment ramp-up cycle. In addition, COVID-19 caused some project cancellations and delays impacted new business activity and data collection projects. This was somewhat offset by growth in China and continued growth in a large transcription project that started in 2019 and continued to ramp up in 2020. Underlying EBITDA of AUD 108.6 million represents an 8% increase over the prior corresponding period. This result was impacted by investment made in sales and marketing in China in order to drive long-term sustainable growth performance. The incremental increase in sales and marketing expense of 50%, an increase in China of 117%, was somewhat offset by strong expense control in the second half of the year.
Other expenses reduced by 11.6% half-on-half, and employee expenses increased 6.9% half-on-half, compared with 37.9% growth for the full year. Management remains committed to prudent management of the cost base and the prioritization of investments that drive future growth and efficiency. As a result, the underlying EBITDA margin of 18.1% was down from 18.8% in the prior corresponding period. Underlying EBITDA includes an FX gain of AUD 6.8 million, comprising a realized gain of AUD 4.7 million from restatement of U.S. dollar- denominated debt drawn to fund Figure Eight earn-out payments. This accounted for most of the first- half FX gain of AUD 3.6 million. It is also comprised of an AUD 2.1 million unrealized gain on the restatement of the hedge book.
Excluding the impact of the FX gain and the investments mentioned previously of AUD 12.7 million, the resultant underlying EBITDA of AUD 114.5 million is up 13% on the prior year result with a margin of 19.1%. Just a very important point of clarification on this FX gain to explain how we're thinking about this because we feel some people may not be thinking about this in the same way or maybe in the correct way. The strong AUD hit our performance in H2 as our revenue was impacted by AUD 15.8 million and our underlying EBITDA by AUD 4.2 million. This impact is included in our EBITDA result of AUD 108.6 million. On the flip side, because of our hedge positions, we were able to achieve a hedging profit of AUD 6.8 million. As we include the negative translation impact, so too should we include the positive hedging impact.
Alternative treatment would be to exclude the negative translation earnings impact of AUD 4.2 million and the FX gain of AUD 3.2 million in the second half, which would see a second- half underlying EBITDA increase by AUD 1 million. Underlying NPAT of AUD 64.4 million represents a 1% decrease on the prior corresponding period. This result was impacted by the after-tax cost of the investments as well as increased amortization resulting from more development work done by more engineers. The effective tax rate for the period has reduced to 20.5% from 24.4%. The effective tax rate is subject to overseas tax rate differential and fluctuations from the tax effect of movements from expensing and vesting of employee performance shares. Excluding these performance share-related movements, the normalized tax rate is circa 28%. Please follow me on page 17. We'll talk a bit more about the currency impact.
We do have a currency impact when we report, that is because almost all revenue and earnings are generated offshore and mainly in U.S. dollars. As a result, we always show the constant currency impact. The full year AU/U.S. rates of AUD 0.6904 was close to our forecast rate of AUD 0.70, resulting in an overall increase to revenue of AUD 6 million and a reduction to underlying EBITDA of AUD 0.6 million. However, the half-on-half swings were very significant. In the first half, FX tailwinds increased revenue by AUD 21.8 million or 7.1% and underlying EBITDA by AUD 3.6 million or 7.3%. As we discussed earlier, in H2, the strong Aussie dollar resulted in reductions to revenue of AUD 15.8 million or 5.4% and to underlying EBITDA of AUD 4.2 million or 7.1%. Over the page onto the balance sheets.
Through solid operating performance and effective working capital management, the balance sheet continues to strengthen. Cash on hand at year-end increased by AUD 3.1 million to AUD 78.4 million. The decrease in trade receivables of AUD 51.7 million should be viewed in conjunction with the increase in contract assets of AUD 33 million as the relevant invoices in respect of completed work at year-end are pending satisfaction of customers' billing milestones or billing periods. The majority of the contract assets were subsequently invoiced on January 1, 2021, and as of February 16, 80% of these invoices have been paid. Receivables also reduced due to delays experienced with customer receipts around the end of 2019, subsequently received in early 2020. Non-current assets comprise mainly goodwill and identifiable intangible assets, mostly arising through acquisition. Following a detailed full-year review, we report significant headroom in the carrying value of these intangibles.
There was no debt at year-end as debt drawn to fund the Figure Eight net payment was repaid in August from cash reserves. Final dividend payment has increased to AUD 0.055, up 10% from the 2019 final dividend, and is franked to 50% . Over to the page to the cash flow. The cash balance of AUD 78.4 million was negatively impacted by the year-end conversion of cash held in USD at strong AUD levels. There was positive impact from the receipt timing at the beginning of the year, as mentioned above. Cash flow from operations is strong and has increased by 39%, driven by effective working capital management. Cash has been effectively deployed for debt repayments, tax, dividends, CapEx, operating expenses, and growth investments. Cash conversion remains strong at 104%. I'll now hand you back to Mark for the rest of the presentation.
Thanks, Kevin. To slide 20. We've made good progress on our ESG initiatives in 2020. Our crowd NPS score is strong at 48, and we're implementing our crowd code of ethics, provides fair pay and open communication and protects privacy. We tackle issues as they rise and enter 2021 with a much improved process than the year before. Our employees were understandably a large focus of our attention this year as we tackled the pandemic together. I'm proud to say that we lived our values, grit in particular, and worked hard, stayed healthy, and saw improved employee engagement despite the challenges of COVID. We continue to focus on material, social, and environmental issues, including our work with the World Economic Forum and Translators Without Borders, and we have released a new environment position statement. In conclusion, the market opportunity for us remains strong.
The chart on page 21 shows ongoing high growth. This is due to the expansion of use cases, projects, and data refresh. We feel fortunate and privileged to contribute to such a dynamic and expanding market. To page 22 and the outlook. It's a uniquely challenging year to provide guidance, as the paucity of guidance from other companies shows, but we're trying to be as transparent as we can despite this. Our order book is solid at AUD 240 million. This is expressed in constant currency to allow comparison with last year. It includes the large, recently re-signed ACV contract also to allow comparison with last year. Excludes, sorry. Underlying EBITDA for the year is expected to be in the range of AUD 120 million-AUD 130 million , again expressed in constant currency, and that's 18%-28% up on last year. In USD, that's $83 million-$90 million.
That's growth in the business of 17%-27%, and we're providing the USD figures to assist year-on-year comparison. Finally, we anticipate EBITDA margins in the high teens. The outlook reflects some near-term uncertainty due to the pace of the economic recovery, the evolving regulatory environment facing our major customers that could necessitate some changes in their priorities. The pace of these changes may weigh on our first- half growth. That said, we're happy to be closely aligned with our customers now and in the future. They are some of the world's most forward-thinking and dynamic businesses, and along with all of our new customers and projects, the AI industry tailwinds our position as the largest player in the market. More committed revenue, our crowd, and our technology all put us in an enviable position of strength for continued growth.
To the final slide, and thank you for your attendance on the call and ongoing support and interest in our business. Before we open the call for questions, I'd like to thank all of my teammates at Appen for their hard work and dedication and for delivering this result to all of you. Now back to the moderator for questions. Thank you.
Thank you. If you wish to ask a question, please press star one on your telephone and wait for your name to be announced. If you wish to cancel your request, please press star two. Please register your question as early as possible. The first question comes from Michael Aspinall with Jefferies. Please go ahead.
Good morning, Mark, Kevin, and Linda. Thanks for taking my questions. Just to start off with, you mentioned that some of the large mature projects that impacted the latter half of FY 2020 are recommencing. Has this started in earnest yet, or is that still to come over the next few months?
Hi, Michael. It's more the latter. There's, as I said, a steady return to these projects. We're monitoring how that improves through the year, but we're absolutely seeing a return to some of that work.
Okay. Yeah, just thinking about the timing in FY 2021 and 2022, that's helpful. Thanks. Some of that impact was customers reprioritizing resources. Do you have a sense of how your large customers are progressing in terms of increasing the total available resources for them?
We do know that our large customers are visibly ramping up technical resources, engineers, and the like, and there's been some public statements to that effect. We also know, though, that companies that used to work very collaboratively and closely in an in-office environment, it's taking a while for systems and processes and cultures and everything to be replicated in the current at-home environment. That is to say that they're actively ramping up resources, but it's taking perhaps a little longer under the current working conditions than hoped.
Okay. Yeah, that makes sense. The new product areas that they're investing in, is it safe to assume that you're working on pilots across the majority of those at the moment?
I don't know all of the things they're working on, it's hard to say what percentage of those things that we're working on. The 34% increase in projects is an indication that we're working in a bunch of new areas.
Okay, just the last one from me. On some of those new product areas for those large customers, can we just think about how they might progress in terms of growing into the larger projects like your more mature ones?
It's hard to know for sure, Michael, because a lot of the growth in data requirement can depend upon the rollout of the product. For example, it might work in one country and they want to roll it out country by country or 10 countries at once, and that changes the data dynamic. However, the nature of some of the products that I'm aware of that we're working on are data heavy. There's potential there, but it's a little early to know for sure.
Okay, great. Thanks.
I think overall, Michael, we're pleased that we're involved in so many new projects rather than just being. Which shows from our customers' perspective that they regard us as an important contributor to their growth rather than just sort of pushing us to one side as they reprioritize. They're actively including us in these developments.
Okay, great. Just the last one from me. You're debt-free again. Are you seeing plenty of options for value-added acquisitions at the moment? How would you prioritize something that may add capability or access to new verticals?
Probably the latter, where we're pretty pleased with the capabilities we've got and particularly the technical capabilities we've got and many of the companies that are highly technical in our space are very early stage and then probably not of interest or probably not financially attractive. Getting into new areas with businesses that are growing well and profitable is the focus for us.
Okay, great. Thanks for that.
Thank you.
Thank you. The next question comes from Garry Sherriff with RBC Capital Markets. Please go ahead.
Yeah. Hi, Mark, Kevin, Belinda, a few questions. The first one, just on your calendar year 2021 EBITDA guidance, that's at spot rates of AUD 0.79. Sorry, it's at AUD 0.69. I just wanted to clarify or confirm the FY 2021 guidance at current spot rates firstly, if that's possible.
Garry, hi. No, we've specifically provided guidance at constant currency and U.S. dollar rates. Essentially, the volatility in the FX makes it very difficult to really draw a comparison. Therefore, Mark pulled out that the U.S. dollar numbers and those growth rates 17%-27% is a really good barometer. Gets rid of all the noise, I guess, in terms of spot and whatever view that's been taken. Spot changes day by day, therefore, the view we've taken just around that guidance is constant currency, but also a barometer in USD to help analysts, I guess, compare what we're saying as a growth rate compared to what growth rates they had in terms of what they had year-over-year for growth.
Okay. Yeah, You seem to have exclude FX gain from underlying EBITDA when calculating the implied calendar year 2021 growth, yet you include it. Yeah, just trying to get some clarity there, but that's fine. We can do that offline.
Thanks for raising that. I think we'd like to address that. We've talked about obviously what the movements were in 2020. We obviously had restatement of performance, obviously at spot, and then we had obviously hedge positions which were contrary to that. When we think about 2021, essentially from a 2021 point of view, there's no kind of assessment in terms of what that future FX is. There's no obviously impact in terms of what that restatement is relative to what was expected. Therefore, in order to get like for like, certainly we haven't assumed anything from an FX gain or loss into 2021. At the same time, we haven't assumed any type of movement from the translation in terms of how the currencies should move.
Essentially, the difference there is in order to get that like for like, because there's no restatement factored in and there's no FX assumption, then we take out the FX because essentially that was only there as a result of, I guess, hedging around the translation. As there's no impact or assumption of the translation in 2021, therefore, there's no room or it's not relevant in terms of having the one-off FX position included in that.
The calendar year 2021 EBITDA margin being in the high teens, could you maybe just clarify, do you expect them to be flat or grow? I guess the second question around that is what sort of first half, second half earning skew should we be thinking about? I did note that you did flag, I guess, it sounds like quite a weak first half and a big second half. Just wanted to get some sort of sense around margins and also that skew, if possible.
Yeah. Hi, Garry. We're planning for margin growth in the year. However, due to the uncertainty we flag in the first half, we still think it'll be in the high teen levels. In terms of the first half, second half split, again, uncertainty in the first half. You will probably see a skew to the second half, but time will tell as we get through the first half as to exactly what that is.
Yeah, I think the point we're calling out, though, is in terms of what you'd normally have seen for us in terms of those splits, this will be quite different in this year for the reasons we've mentioned.
Okay. The last question, just in terms of the contents relevant revenue, do you know what portion of that is advertising related? I imagine the vast majority, but I just wanted to clarify that with us.
It's substantial. We don't split it out like that, but it is a substantial portion.
Are you, I guess, looking forward back in December, is it fair to assume that now when you look at advertising- related revenue growth or the profile growth, do you think that's now materially different or more permanent maybe might be a bit more accurate? Should we be thinking that the growth path for content relevance growth is maybe lower from advertising revenue specifically? Any guide on that would be interesting.
Yeah, it's a good question, and it's one that we ask, and hence flagging uncertainty in the first half. The thing that's impacting it, Garry, is the speed with which the customers sort of set their overall product priority. Do they go hard into these new areas and potentially under-invest in current areas like advertising? Or do they ramp them both up at a similar pace? Or do they moderate investment in the new areas and go hard in advertising? That's the nature of the uncertainty that we're flagging. It's the customer's decisions around how they invest in their product portfolio, which is giving us some pause for the first half.
Last one, just around the change for the identifiers for advertisers that Apple's proposing for their next iOS update, effectively allowing users to opt out or not share their data. Interested to get your views as to how you think this could affect your customers' ability to generate ad revenue and how it might impact you guys once those updates start to roll through in the near future.
Yeah, interesting question. Clearly, our customers benefit from a lot of data that enables them to build highly personalized ad targeting engines. If that data source or if a source of data is throttled in some way, they will either have to change the way they do things or look for another source of data. Some of the projects that we work on is about providing representative data sets of different demographics. For example, our customer may want some enhanced relevance work in a particular country amongst a particular age group of people, and we provide that for them. Overall, this is an interesting area that will cause or may cause or may provide some opportunities for us, if that makes sense, as the customer looks for sources of representative data, if they can't get highly personalized data due to changes in technology.
Okay, thanks. I'll step back in the queue. Thank you.
Thank you. The next question comes from Siraj Ahmed with Citi. Please go ahead.
Thanks. Few questions. Mark, just first thing, if you just break down the growth into price and volume, just keen to understand if you're seeing any pricing or price per data point pressures.
Hey, Siraj. Not overly. It's, I would say, regular pricing pressure. Through the year, we have an agreement with one of our major customers that puts a floor on pricing. We also have strategic agreements with our other customers that keep pricing where it is. Where we do see pricing pressure is in some of the new work and the new customers that we do. It falls into two camps. It's kind of there or thereabouts as to where our pricing is, or sometimes it's just wildly different, and we don't understand how that pricing comes about. Those instances, they're very few and very small projects. Overall, there's no sort of material price pressure in the business.
Got it. Secondly, just on the work in hand number, can you just clarify, did you say that it does not include the committed ACV?
We took that out because it messes up the like-to-like comparison. It's the same methodology. The order book is comprised in the same way that we did it last year.
Yeah, that contract was basically signed after we announced in February last year. Like for like, taken out.
Got it. Yeah, okay. Like for like. In the year-to-date revenue number, there will be some pickup in work, but the work in hand number, you haven't put the ACV in. Is that a way to think about it?
The large ACV contract has got a February date on it, so it doesn't feature in the January year-to-date number.
Got it. Maybe just trying to understand this, because if you think on constant currency terms, the number has grown 14%, right, year-on-year. I think the second half growth in constant currency was 6%. It looks like you're seeing a pickup in growth. Is that fair? Where is it coming from? Are you assuming the same fourth quarter skew this year in your guidance?
Sorry, Siraj, didn't hear the last bit. Just to answer your question, yeah. What you're talking about is increase in the order book, 14% in constant currency, correct. I would just caution you, obviously, that's orders received at a point in time for the rest of the year, and obviously the timing of receiving those orders could have a dramatic impact. It's a guide, but you need to just understand things that can impact that. Sorry, can you just repeat that last part?
Yeah, no, I heard it. In regard to the Q4 skew, Siraj, I think what we're calling out this year is a period of uncertainty in the first half. That will most likely cause a skew to the second half. We're sort of moderating our view on Q4, given what we experienced in 2020. There's definitely a half-on-half skew due to some uncertainty in the first half.
Got it. Just last one from me. Just looking at the cohort chart that you put in slide eight, Mark. It looks like the new customer you had from Leapforce, that had a bit of slowdown, looking at that chart. Can you just talk to that? The dark red one continues to grow, but the other one has slowed down a bit from the looks of it.
Yeah. It goes to the general theme last year, Siraj, of advertising revenue or advertising-related programs being impacted by the slowdown in ad revenue. As I said earlier, on present indication, there's a steady return to most of those programs. You're correct in that customer was impacted, and it's because of that general ad-related theme.
All right, great. Thanks. I'll jump back in with you. Thanks.
Thank you.
Thank you. The next question comes from Lucy Huang with Bank of America. Please go ahead.
Good morning, Mark and Kevin. Thanks for taking questions. I just have two. Firstly, in the speech and image division, the revenues have declined. Just wondering whether you're seeing this. I know you noted that there's just a bit more cancellation of orders and deferral of projects, but what do you think on the competition front? Has there been any change in any of the dynamics there? Just secondly, is there any way to extract further efficiencies from the business to kind of over time potentially lift that mid-teen EBITDA margin a bit higher? Just wondering whether there is that aspiration and maybe what kind of initiatives or investments are in place to try and achieve that. Thanks.
Yeah. Hey, Lucy. In terms of competition, there's no material change in the competitive landscape in terms of the companies we see. I know there's been some attention on that recently, but we've been monitoring these companies for some time. Some of them have picked up funding recently, but that builds on other funding they've got and in some ways, that validates the market we're in. With competition, it's also helpful to break it out into the various data lines that we work in. Most of the competitors are in the image space. That's the, I guess, the easiest problem to solve. It's also an attractive area for new entrants because of things like autonomous vehicles, et cetera. Speech and natural language has far fewer competitors. It's a much more specialized space.
Some of the competitors that are getting all the press just don't feature in that space and/or struggle in that space. Then to relevance, which is the most specialized overall because of the scale of the operation, and the competitive dynamic remains the same with us and our major competitor. Overall, there's no sort of material shift in the name and number of competitors in the space. To efficiencies, yes, we can absolutely extract more out of the business, and that's the plan, to lift margins permanently beyond the teens. We ran into a small problem called COVID in 2020 that necessitated us to sort of hunker down on some of our major change programs and focus on our staff and our crowd and our customers. We haven't progressed some of those initiatives to the extent that we had hoped.
They're underway, and we're going to double our efforts into 2021 to grow those margins.
Yeah. I think the other point just to note that's important here is that given that the revenue can move around is largely tied to product and investment life cycles. It's not really that we lose customers, it's just that we need to wait for the next cycle to come through. As a result of that, we don't take significant dynamic changes to our expense base as a result of that. What that means is that you actually have a lot of sensitivity in terms of revenue to bottom line. Less revenue with largely unadjusted base. That's not to say we don't look at it, but we need to be aware that because we're positioned well for the next cycle, that we need to have a support base.
That will obviously still be a big factor in terms of what exactly those margins are on a period-to-period basis.
Wonderful. Thanks. I might just follow up with one last question. I think you mentioned image is quite competitive, lots of new entrants, whereas speech has fewer competitors. Just wondering, will there be incremental investment in image and speech? Are there some capability gaps in those two areas which may need to be plugged over time? Do you think the focus will shift away from image into speech or just more into relevance more broadly?
I think a focus area of capability for us is in automating the speech and image areas with AI. I mentioned a couple of examples in the presentation, and they're spelled out there for your reference. The more we can automate the work using AI, and that's not just the actual work, but the preparation of the data, et cetera, the more we can win, the more we can do, the better our margins, the higher our revenue. The focus, and I wouldn't call it a gap in the sense that our competitors are ahead of us in this area. It's definitely an area that we're focusing on to enable greater provision of data at higher speeds and higher quality for our customers.
Wonderful. Thank you.
Thank you. The next question comes from Bob Chen with JP Morgan. Please go ahead.
Hey. Morning, guys. Just a few questions from me. I guess, looking at sort of the full-year guidance there, and you sort of calling out some first-half uncertainty. I mean, what gives you that comfort to provide that full-year guidance? Are these discussions that you're having with your key customers that's providing you a better look into the pipeline into the second-half?
Hey, Bob. The order book includes work that is to be delivered this year. The customer is calling out some uncertainty in the first half, but they're placing the orders. That gives us some confidence that revenue's there and will be delivered. I'll also point out that all the way through the pandemic, our at-home crowd model continued to deliver into the customer requirements. We don't have a supply side problem at all. Yes, it is conversations with the customers. Obviously, we've been very closely in touch with them throughout the year. It's also the fact that we've got many more customers. It's also the fact that we've got many more projects. There's a lot of optimism in the system, and that goes to confidence in the forecast. Notwithstanding just the general trend around AI continues to be positive.
Yeah. Okay. Then, looking at some of the U.S. results versus Q4 advertising results, I mean, they came out pretty strongly. How do we sort of reconcile that to the softness that you guys saw into Q4? I mean, is there typically a bit of a delay in the planning from the key customers? How does that all work?
A lag is part of it, but it also goes back to this point that we're making around the uncertainty in the first half as our customers look at their product development portfolio. Keep in mind, the history of a lot of the large tech players has been, for want of a better word, unbridled growth, right? They've just gone ahead and they've done what they needed to do. The pandemic was a material sort of event for these companies. In doing so, they're looking at their product strategies and what they're doing going forward. That's not to also discount the regulatory environment that they face. They've got some headwinds to navigate. Having said that, they're very smart. They're very dynamic. They're very clever and resilient companies, and they'll find solutions, and we're part of helping them find solutions for these things.
Okay, great. Just touching on the cost base, obviously called out that AUD 10 million investment in sales and marketing over FY 2020, what's the outlook into 2021? Did you sort of expect another step up in sales and marketing, or are you going to see how your existing sales and marketing investment goes?
Yeah, it's the latter, Bob. We'll see the cost base normalize through 2021. We go into the year with a full cost base, which is another weight on the first half result. We don't anticipate, nor have we budgeted for any big step-ups in investment in areas like sales and marketing or technology. Just to remind you, a lot of investment in tech in 2019 and then a lot of investment in sales and marketing in 2020 and more of a normalized spending pattern through 2021.
All right. Thanks, guys.
Thank you. The next question comes from Quinn Pearson with Credit Suisse. Please go ahead.
Hi. Good morning. I was hoping to talk a little about new customer acquisition. You have a chart in there showing that there's some growth there, which is great to see. I was hoping you could talk us through your sales and marketing efforts and how you feel those are progressing and how you feel that sales and marketing program to new customers is evolving. This is obviously a very different go-to-market strategy than the historical highly concentrated customer base. I guess any learnings, if you feel like you're making good traction in progress, and if you think we should be expecting any kind of acceleration in new customer acquisition or if we're finding that more challenging than had previously thought. Thanks.
Hey, Quinn. Some success in 2020 with new business development and a lot of lessons. We did a review of our go-to-market work early in this year. Some of the findings that we're starting to implement were quite interesting. First of all, the customers highly value our expertise and our scale, and we need to make more of that. Our tagline of confidence to deploy AI in the real world speaks to that. If you want to deploy your AI and not have any problems in the real world, then we have the experience and the scale to do that. We need to leverage that message a lot more. We also need to ensure that we have the right technical expertise in front of our customers, because that's what wins the day.
When we put our project managers, for example, or our linguists that have actually done work similar to the customer or similar to what the customer wants to do, that gives the customer an enormous amount of confidence. We found that selling outside of the major tech companies required more of this technical expertise. The enterprise customers, for want of a better word, not the technology customers, didn't understand how to prepare data sets as well as our major customers. That added a bit of a burden to the sales motion. Finally, in going back to the point on price, and our go-to-market study included a lot of customer interviews. We found that our customers said pricing is an issue. Quality and suitability of data. Quality and data accuracy is what's the most important. Price is a factor, but it's not the most important factor.
A lot of good lessons, and we'll leverage those into 2021. I anticipate we'll maintain our trajectory of new customer wins. At the same time, we're going to grow the projects that we've won and increase the revenue out of those new customers as well.
That's helpful color. Thanks. Just to confirm, did I hear you correctly that some of these new customers you're seeing large variances in pricing where some is in line with where you've previously expected and some outliers are much lower? Was that correct with new customers, or was that more just with new projects with existing customers?
No, that was definitely the new customers. We couldn't see a pattern in the outliers. For example, we'd bid on a project, and if there was a conversation around price, it was single-digit percentage points. We'd have a customer saying, "Well, we got a price on this that's one-tenth of what you're telling us." We're like, "Well, how does that work?" In those situations, they were small projects. They could have been exploratory. It seems just too random to be a sort of a genuine opportunity. In response to questions on price, I guess I'm just being transparent in what we're hearing, and we hear two fairly polarized things. Most of the time we're on market, and then occasionally there's these absolute bizarre-type things.
That's helpful color and transparency. In general, you've been transitioning the business more towards a scalable platform. I guess what I'm hearing is it sounds like some of these new customers don't have quite the same types of internal capabilities. Is there a business case to almost pivot the other way and start adding more consultant-type services to help support some of these new customer opportunities?
Yeah. Fair question. Although we see it slightly differently and maybe I need to give a little extra color. If we look at business development outside of the major tech players. Amongst the big five tech players, we absolutely take a very collaborative, very consultative approach because their programs can be very big, and it's up to us to do what the customer needs to do. Outside of that, we still see a lot of variability in the nature of the projects, but we have the opportunity to coach the customer along a more replicable line, if you will. I can't think of a good analogy off the top of my head.
If you've ever walked the aisles of Bunnings trying to work out how to do something at home, and then you meet a tradesman and they can do the same thing but in a different way, but they do it every day. We bring that sort of pragmatic expertise to the customer where we can achieve the same outcome for them, but perhaps in a different way to the way they were thinking about it. The value in us bringing our expertise and our technology platform to the customer is they can actually see and understand, "Aha, that's how I get the outcome that I was after. That makes perfect sense." There is a repeatable model there.
There's a lot of variety in the use cases that we tackle, but amongst people that are earlier in their journeys, there's the opportunity to coach them down a more repeatable way of doing things.
That's helpful. Thanks. Lastly from me, you provided some helpful color on the competitive set across the different categories of work, relevance, speech, image. I was hoping you could also put an overlay of customer dynamics and end customer demand across those and perhaps put in order where we should be looking for growth to be led at a group level versus where it might be growing slower than the group across the relevance, speech, and image, please.
I think the essential dynamic there, Quinn, is as we called out between relevance and speech and image. Relevance has that ongoing refresh requirement which will go to ongoing growth. Yes, we know that was lower than it has been historically last year, but we're starting to deal in some very big numbers with the relevance revenue. The speech and image relies on us winning more and more customers to get higher growth rates because the nature of the projects, they're more cyclical in terms of the data requirements. We have to do three things very well. We have to look after our big customers and be there to support them on their existing and new projects. We have to support and grow all of the new customers that we've won in 2020, and then we have to continue to win new customers.
I'm confident in our ability to do all of those things. We've got a well-established sales and marketing capability now that delivered a lot of new customers in 2020, and it'll continue to do that into 2021. We've got to do all those things to get continued growth.
Helpful color. Thanks for your time.
Thank you.
Thank you. The next question comes from Paul Mason with Evans & Partners. Please go ahead.
Hey, guys. Just a couple from me. The first one's to do with your platform CapEx. Two questions related. The first is your trajectory looks like it's stepped up in the second half versus the first half in terms of your capitalized spend. I was just wondering, so should we think about the second half run rate as sort of what you're going to do in 2021? Should we think about the annualized sort of total as the amount you're going to do in 2021? Related to that, could you maybe make a comment on whether any of those automated annotation tools are actually in production now or whether they're all still in the development phase? I've got some other follow-ons after that.
Yeah. Hi, Paul. I'll handle the first part of it. To guide you guys, you should be thinking about H2 as a proxy for 2021. H2 annualization for 2021 proxy.
On the automation, Paul, some of the automation tools are in production, but in very early stages. One thing we absolutely don't want to do is mess up the quality of the data and service that we provide for our customers. So we're making sure that we start small in the use of the automation tools. I can give one example, which is for one of our customers that we do a lot of transcription work for. We are using AI to pull apart the data. Specifically, we get a page of text, a scan of an image of a page of text. The annotator highlights the whole line of text, and then the AI automatically breaks it into words and identifies the gaps and the punctuations, which speeds up the transcription work, which is the next part of the process.
It might sound fairly trivial, but it means that the annotator is just saying, "Okay, split up this text," and then it's straight into the transcription queue as opposed to identifying each word in the workflow. The summary is generally more pilot than production at this point. We see some productivity benefits as we roll those more aggressively into production.
Okay, great. Just a question related to, you guys have a customer liabilities balance, which has been shrinking, and I think that that's related to Figure Eight used to pay for a third-party annotation service, and now that it's internalized, you're sort of not requiring as much upfront payment as what Figure Eight used to. Could you maybe comment, should we expect that balance to basically go to zero over time? Or has it sort of hit its low point now in terms of the levels of the customer liabilities?
Paul, I think when it comes to the customer deposit side of things, yes. That's very much tied to the migration of the traditional Figure Eight third-party providers who used to do the labeling work now being done by us. Obviously the end goal is that we do all that work. Once that happens, then from that point of view, that could be the bottom point for customer deposits. The rest of the movement is along the lines of the deferred revenue to do with the contracts, and obviously that will still be very much a feature.
Yep. Okay. Just the last one from me. Just obviously Lionbridge AI transacted during the period, and I was just wondering if you could make any comments on whether you participated in the process at all, or if you didn't, why you didn't?
I think we said previously, Paul, that our customers like the fact that they've got a couple of vendors to choose from, and I think it would be strategically unwise for us to look at something like that.
Okay. All right. I'll leave it there. Thank you for your time.
Thank you. Our next question comes from Raimond Desdi Cerappos with PineBridge Investments. Please go ahead.
Hi, Mark. Hi, Kevin. Can you hear me?
Yes, we can. How are you?
I am well, thanks. Hope you're well as well. Just a question from me on the large projects recommencing. You noted some weakness in these projects late last year, I think the big question then was whether it's a temporary reprioritization, as you said, or a structural change in the customer's view towards the data requirements for these more mature projects. At that time, you sort of believed that it's a former rather than the latter, but sort of mentioned the uncertainty due to your lack of understanding into the inner workings of the customers.
Just wondering, with the comments now that you are seeing these projects recommencing and some conversations with the customers in between, can you give us an updated perspective or even a level of confidence on whether or not that rollout that you saw and are seeing now is indeed temporary versus structural in nature? Thank you.
It appears to be much more temporary than structural, Raimond. However, the pace of the return is the thing that we're calling out as uncertain, per the other responses. Definitely it's looking far more temporary than structural, but the pace of it getting back to where it was is the bit that's uncertain at this point.
Mm-hmm. Okay. A question on the order book figure, that $240 million. It looks quite solid to me. As a percentage of the implied sales, given the EBITDA margins that we expect and the EBITDA number, it looks to be a percentage that is higher or at least in line with the historical trends. Just want to reconcile this seemingly strong order book with the comments of sort of uncertainty from the customer side and a customer wait-and-see attitude. Really, how does this strong order book number reconcile with those softer comments on your side?
I think it can be consistent, first of all, because the order book's for the full year, and we're calling estimates certainly in the first half.
There's an implied skew to the second half in activity, and that has come directly from conversations with customers. Our customers are just being a little more considered in the way they're getting back to these programs because they've got a lot of considerations to consider. That's not very good, is it? There's a lot of things to think about. They've got the pace of the return to business as usual under COVID-19, and things are looking good, but the world turned upside down 12 months ago, and that's maybe weighing on their minds. Certainly weighs on mine. You've got the regulatory environment that they're facing.
They've got a bit more of a complex world to deal with at this point, and they're trying to work through that, and they're calling that out in their work with us, which I think is good that they're being very transparent with us. Summary is, the order book's the full year. We're calling out some uncertainty in the first half, and as we progress through the first half, we'll keep the market informed as things evolve.
Okay. Thank you. Last question from me on those newer, less data-intensive projects from the large customers, those new initiatives. From having worked on these projects and understanding their nature, can these projects ever be as large as the more mature ones that are seeing weakness now? What would be their ramp-up profile? Can we expect them to achieve substantial scale, say, this year or the next, or would this have to be a sort of multi-year kind of wait?
It's a little early to tell, Raimond. Knowing what we know about these programs, they're certainly data rich, and that could lead to a substantial ramp-up. It depends very much on the success of the product in the market. You've got existing programs that have got tremendous market traction and good market position that need an amount of data to stay relevant. Then you've got a bunch of new product developments that are unproven and should they all take off, it could be terrific. Again, it's just a little early in their life cycle to know.
Okay. Thank you. That's all from me.
Thank you.
Thank you.
I'm thoroughly sorry, but we're going to have to wrap the call up. I have another meeting in a minute's time, and we're 15 minutes over. We very much enjoyed the questions. We look forward to talking to people one-to-one through the week, but I'll hand it back to the moderator to close the call.
Thank you. That does conclude our conference for today. Thank you for participating. You may now disconnect.