Appen Limited (APX) Earnings Call Transcript & Summary

Ladies and gentlemen, thank you for standing by, and welcome to the Appen Limited Investor Update. [Operator Instructions] Please be advised that today's conference is being recorded. And I'd now like to hand the conference over to your first speaker today, CEO, Mr. Mark Brayan. Thank you. Please go ahead.

Thank you very much. Hello, everybody. Welcome to the trading update for Appen on the 10th of December 2020. My name is Mark Brayan. I'm the CEO of the company, and I'm joined here this morning by Kevin Levine, our CFO; and Linda Carroll, our Head of Investor Relations. We'll stick through the release and then allocate some time for questions. Some background to begin with. You'll recall that we identified a number of possible headwinds at the outset of the pandemic. We thought that it might impact our performance through a slowdown in digital ad spending, which, by the way, is the dominant source of revenue for some of our largest customers and a good indicator of their spend with us. We also thought that we might see a reduction in IT spending in general, a reduction or cancellation of services from some of our smallest customers, interruptions to global hardware supply chains and suspension of face-to-face work, such as audio data collection. Now we reinforced this at the half when we reported slower new sales than hoped and lower renewals. We also called out that we expected the slowdown in online advertising spend to have a negative impact on our ad-related programs in the second half. At the time, we thought this would be a small impact based on what we were seeing in the flow of projects and sales. Our Q3 revenue was lower than expected, but the quarter came with some good news. Our major customers released strong Q3 results and online advertising bounced back. When we reviewed the projects in our pipeline, took into account that our Q4 revenue historically averages 30% of the full year and enacted some cost control measures, we were confident that we would achieve or be very close to our guidance range and hence, maintain guidance. However, we've just finished our November results. And while Q4 has improved on Q3, work that we're anticipating in November and December has not yet materialized, and we now expect full year '20 underlying EBITDA, including the impact of the stronger Australian dollar, to be in the range of $106.0 million to $109.0 million, and that's at actual exchange rates to November and $0.74 for December. If we apply our originally assumed exchange rate of $0.70 for the second half, this is equivalent to a guidance range of $108 million to $111 million. This means that second half underlying EBITDA for the year is expected to grow at 30% plus over the first half, applying first half exchange rates to second half performance. So COVID is clearly having a greater impact on our business than anticipated. While the long-term trends for AI post COVID are positive, and there are a number of things happening currently. So first of all, COVID has disrupted the way that our customers work. We deal with engineering teams and their creativity and problem-solving thrives on teamwork, and that's been upended during the pandemic. Now while our at-home crowd delivery model has been resilient, our face-to-face sales and customer engagement practices have been severely impacted. Despite what you read in the paper, it turns out that it's really hard running a high-growth business over Zoom. Keep in mind that Australia is enjoying a good pandemic, if such a thing exists. Cases in the U.S. are rising. In California, the home of our biggest customers, has recently announced more intense lockdowns, further exacerbating the challenges. At the same time, the pandemic has meant that our major customers are accelerating their new product development. Winning products enhance their long-term resilience and value and the more winning products they have, the better. For example, earlier in the pandemic, Facebook announced that they would hire 10,000 people by the end of 2020 in their product and engineering teams to deliver on such a strategy. And across our major customers, the number of projects we're working on has increased 32% from 2019 to 2020. This is extremely positive for us. But we are currently seeing customers' resources reprioritized towards these new projects and away from some more established projects. The material projects have slowed as a result and reduced our revenue and the new projects are yet to require the data volumes that offset the slowdown. So although this is impacting this year, it is setting a foundation for a strong year in 2021. In addition, we continue to win new customers in markets less impacted by COVID, including in areas such as shipping, automotive, education and health and the long-term trends for our business remain positive. Spending on artificial intelligence is growing rapidly, 28% annually, and AI adoption should accelerate in a post-pandemic environment. Further, online advertising, remember, a major source of revenue for our customers and a reasonable indicator of their spend with us, is forecast to rebound strongly in 2021 according to financial analysts. Our pipeline for 2021 is building nicely, and we look forward to updating you with a strong order book when we release our results in late February. These structural tailwinds as well as the strength of our existing pipeline for '21 support a return to strong growth rates in '21, in line with industry trends. Thanks very much for joining the call and your attendance, and I'll now hand it back to the moderator to take your questions. Thank you.

[Operator Instructions] Our first question is from Lucy Huang from Bank of America.

So I have two. So firstly, you mentioned that you're seeing a strong order book going into 2021. I'm just wondering how much visibility do you have and how much of that spend for 2021 is now committed? And then secondly, you mentioned the tech giants are doing a lot of new product development, and so there is an opportunity there. I guess how confident is Appen in terms of winning their fair share of work across the new projects that are coming ahead?

Yes. Thanks, Lucy. So in terms of the visibility, I'd say we've got more visibility than we have had in prior years. However, we're still working on those numbers. We've yet to receive actual orders for next year, but we're in close contact with the customer, and we're getting a greater level of visibility. It's a little too early to start talking numbers, but certainly, the indications are more positive going into '21 than going into '20 when -- which was prepandemic. And in terms of committed revenue, again, a little early to tell on that one as well. But again, we're confident we'll get a good slice of committed revenue in '21 also. In terms of new product work, look, the good news is we're already on those projects. And some of them are in fairly exciting areas, and some of them are quite diverse compared to other projects that we've traditionally worked on. So I think we're going to -- I think we are well positioned to benefit from that new product development. But they are early stage. We're talking fractions of the volumes of data compared to our largest projects. And that obviously impacts the revenue that we derive from them. But we're in them, and we've got a terrific foundation for '21.

Our next telephone question is from Garry Sherriff from RBC.

A couple of questions. The first one, just trying to pull apart you saying that the larger customers -- existing large customers are spending less with you. Can you maybe quantify where those lower dollars are coming from? Is it speech and image projects or is it Content Relevance and Figure Eight or is it across all 3? That's the first question.

So Garry, sorry, just to be clear, they're not spending less with us. The spend for our large customers is generally growing. Kevin's nodding. However, what we had anticipated was every December, we get a strong surge of work we have done over the last few years. And we didn't get any indication that, that strong surge wouldn't be there. Unfortunately, it's not there to the extent that we anticipated. So the first point is, the revenue isn't going down. It's just not what we anticipated running into the end of the year. In terms of your second question, it relates mainly to some of the larger Relevance programs. There's ups and downs across the board, but it's some of those larger Relevance programs that we haven't seen the volumes of work that we anticipated. And it is because of that reason, as I said, the -- we're required to work with engineering teams who demand the data and use the data, and there just aren't as many engineers on those teams. They're being reprioritized to other areas. So it's not -- we think it's a sort of almost we're in the middle of a bit of a storm of activity where they're spreading their resources out, which has meant that some areas aren't getting the attention that is required, and that's impacting our work volumes and our revenue. So in summary -- sorry, long response. The work volumes aren't reducing with our major customers. We just didn't get a surge in December that we anticipated, and it is mainly from some of the larger Relevance programs.

Okay. Second question, do you think it's customer behavior or is it your share that you think that might also be falling, i.e., do you know there's competition heating up? Are you aware of the competitors potentially taking some of this volume from you? Or do you believe it is purely a customer behavior issue in terms of that November onwards surge that you usually receive?

Yes. It is, in our view, almost certainly customer behavior. We -- the competitive dynamics outside of our major competitor, the Lionbridge, the other competitive dynamics that are in the market tend to be outside of the major customers and definitely outside of the major Relevance programs. So our view is this is customer behavior, we're not losing share. And we keep our eye on that, Garry, through things like the hiring that our competitors are doing. They crowd higher rates and so on and so forth. So we had some indicators of that. The other thing that impacts the amount of work that we get from our customers is whether we're achieving our quality goals. And we certainly are. In fact, for the majority of these programs, we're exceeding quality levels, which is due to the resilience of the at-home crowd model, it's just that there's less work overall. We're doing a good job into the areas that we're being asked to work, but it's just a little bit less than we anticipated in December.

Our next telephone question is from Michael Aspinall from Jefferies.

A couple from me. You mentioned those larger and mature projects in Relevance. If you think about the types of projects, would you expect that to return? Or would you expect that those projects, in particular, spend would just remain lower?

It's a little early to know for sure, Michael. The nature of the Relevance projects is it is refreshing data. And so there's -- risk is the data isn't refreshed frequently enough that the model [ might ] risk, et cetera. But also, there's -- the customer can make that decision. They may accept a bit of that risk to reallocate resources in other areas. So structurally, we assume that it would come back at some point, but it's a little early to tell because we don't know sort of all of the -- all the story there.

Yes. Okay. But structurally, is it something that you would expect to come back, but can't say with 100% certainty?

I'd say so, yes.

Yes. And so FY '21, you mentioned it's expected to grow at industry growth rates. Would you expect the industry, just given what you're seeing, to rebound in FY '21 off of the COVID-impacted year? Or would you expect something more akin to return to trend growth off a lower base?

So we had a view at the half and we maintain that, that once the world sort of gets through COVID, businesses that thrive and survive will focus more on digitization for resilience and scalability and part of that digitization will definitely be artificial intelligence. So from a distance, there's no reason to bet against AI post pandemic. I think the thing that we found certainly over the last month or so is that, it's just when that rebound occurs. Given just the simple fact that you need engineers to build products and if they're all working out of their homes, they're just not as effective. So longer term, we're still extraordinarily bullish on the market and where we are and what we're doing. But it's just the timing of the rebound is a bit unclear.

[Operator Instructions] Our next telephone question is from Josh Kannourakis from UBS.

Just first off, just wanted to chat about in terms of that reprioritization. Just when we're thinking about those mature projects, is how much do you think is resource related in terms of those businesses that have been impacted by COVID in terms of resourcing that they can put on to those projects or mature projects versus potential near-term budgetary restraints?

Hey, Josh, I don't know if I quite understand the question.

I think what I was trying to say is, obviously, there's been lockdowns in California, and those guys have been impacted. People that are working on these projects have obviously been impacted. When you're talking and having the conversations about this reprioritization to new projects versus mature projects, has there actually been a resourcing reallocation, not just to you, but in terms of their underlying businesses away from these projects because of COVID? Or do you think it's more so strategic or more so budgetary in nature?

Yes. Okay. I think it's -- if I could use the expression, accelerated strategy, right? So if you're in the business of building products, you want to keep a steady cadence of new products coming into the market. This is my sort of view from afar, and I haven't had conversations to this extent. But faced with something like a pandemic, these are opportunities for the digital giants. So the more products they have, the more they can benefit post pandemic. So it's -- the vibe we get is, let's get our skates on and let's build as many products as we can to make our businesses more resilient post pandemic. And I think just the speed with which they're doing that has led to this resource choke hold. Now whether the resource allocation becomes permanent, which is kind of Michael's question previously, it's a little early to tell. But certainly, that's, to our mind, the dominant impact on our sort of near-term results. They've just [ reached ] resources away from major projects that require a lot of data to newer projects that are yet to be as data intense and that's had a direct impact on our business. We view it positively longer-term because there's more projects that we're involved in. And history shows, as you well know, Josh, that we get entrenched in these projects, and they can grow quite nicely. But that ultimate resource allocation is still a little unclear, but we like to think that being involved in more product projects at a simple level is a good thing.

Got it. And then just as a follow-on to that, when you're looking, you talked about having greater visibility rolling into next year. Can we talk a little bit about how your efforts are going with customers to get minimum commitments and sort of have a conversation about partnering in that respect more?

Yes. So I think we reported at the half that the smaller customers and the renewals were being impacted by COVID because of that sort of face-to-face selling motion. And that has played through into the second half. So I think it's fair to say that we're not where we wanted to be on new customer wins and committed revenue and other things, but we continue to win new customers, we're winning quite a few of them, they're fairly small, but we're not kind of where we want to be. I think in general, we read a lot in the paper about how productive it is working from home. I tend to disagree with that wholeheartedly. And you and I know, Josh, the interactions that we've had over the years, certainly, they're far more productive face-to-face. And I think that's a big impact here on both our side and the customer side.

Got it. I just wanted to ask on the Lionbridge AI, the acquisition by TELUS, just interested in any sort of market views or initial impressions on that, Mark or Kevin? That would be really interesting as well.

Yes. What I have heard back from some folks is that, generally, they view it as positive for us. I think the release disclosed some numbers that, frankly, I thought they were a little bit bigger than they reported. So perhaps we are outpacing them in some areas. I think also, there'll be a distraction as there always is in those sort of situations dejecting an acquisition, et cetera. But they do a good job. We respect them as a competitor, and we'll continue to invest in our technology and sales and marketing and onwards to make sure the we stay ahead.

Our next telephone question is from Siraj Ahmed from Citigroup.

Just first question, Mark, you mentioned you had cost control measures, can you just expand on that? Because clearly, at the start of the year, you're expanding your sales and marketing teams to go after new opportunities. Just where you are with that.

Yes. So what we did, Siraj, is we pushed costs that we could later into the year and some of it into '21. So nothing that was directly related to revenue. We also didn't materially impact our engineering hiring because we wanted to make sure that we continue to invest into the technology. So in some peripheral areas, some of the new sales heads that we had scheduled for the second half, we delayed hiring them. We wanted to make sure that existing folks we've got are highly productive. So we took measures where we could. As I said, however, we've onboarded over 300 people, and many of those in China and the Philippines, which, of course, are at a lower cost point, but over 100 of those were in the U.S. So we continue to invest into the business to respond to growth, as I think we should. But we did exercise as much discretion as we could to delay hiring and other costs just to be prudent.

Mark, just on that, would it be fair to assume that you have the OpEx basis at the right size for next year [ heading in 2020 the big step-up ]?

Yes. Largely, there's a lot of cost that's directly related to revenue. So we will add some additional resource to deliver on growth. However, there's a lot of infrastructure in place that is leverageable. We continue to invest into the government and Chinese markets. So they put a bit of a drag on margins. But outside of that, into the sort of the core business, yes, we should see some leverage coming through next year.

Got it. And just can I -- can you give us an update on the government as well, please, just to what's going on there? Is the delay because of the elections and stuff, any more there?

Yes. So yes, our government team got hit with a triple-whammy, not only the pandemic and an election, but an election and then some this year. So progress has been slower than we would have hoped. Ordinarily, what happens is in the transition of government, there are things called continuing resolutions that maintain funding for projects. Those resolutions and that whole transition process has been somewhat novel this year, which has affected some of their work. But that aside, longer term, it's still a positive place for us. And relatively speaking, our investments are modest, and we continue to invest into that market. But no big news out of government at the moment.

And next telephone question is from Bob Chen from JPMorgan.

First question for me, just on that talk about some of the reprioritization from your key customers and these newer projects that they're working on. I mean it's too early days now, but do you get any sense as you're working on some of these projects, the volume of data that might be required when these projects become more mature and how that might compare with sort of Relevance projects to date?

Yes. Thanks, Bob. Look, it's a little early to tell. Some of them appear to be very data-intensive. The applications are image-based, for example, and there's infinite variety in the particular use case. So that appears to be fairly data-heavy. The Relevance projects have a very high refresh rate as well, which is why there's been that sort of impact that we're seeing. But it's just a little early to tell as to what the cumulative effect will be of all of the new projects as they ramp up and whether the slowdown that we've seen on some projects, it's not all projects by any stretch, but whether the slowdown we've seen on some projects persists and to what extent. But overall, we think that the number of new projects that we're working on and the commitment that the customers have to developing new products is net positive despite this sort of short-term pain.

Okay. Great. And then just in terms of -- on the cost side of the equation. Given the sort of slowdown into your core mature business in Q3, Q4, do you sort of plan to make any changes to how you approach investment over the next couple of years?

So again, just to be clear, the trend through the second half was reasonably as expected. And we just saw a -- we didn't get the surge we were expecting in November, December. In terms of investments, there's sort of operating expense to support the customers, as I touched on in a prior answer, investing into China and the government markets, which will be a little drag on margins, continue to invest in technology because we still view -- we're still very strong supporters of the thesis that the technology is what's going to accelerate and ensure scalability of our business. We don't have any sort of extraordinary investments in sight at the moment. It's all stuff that we have talked about before. And in general, we're going to persist with that cadence of spend.

And next telephone question is from Wei Sim from Macquarie.

My question is just related to, previously, you mentioned that the work -- or the fourth quarter that we've seen hasn't been as strong, and this has been due to a work that hasn't materialized or not existed or it's been delayed to 2021. I'm just wondering, do we have a sense as to how much of this may have actually been delayed to 2021 in terms of the downgrading guidance that we've given today, how much of that could actually be flowing through to next year? And then following on from that, how would this impact on our outlook for EBITDA margins in 2021?

Yes. Thanks, Wei. To the first question, some of the -- the work that we didn't see in November and December is -- has been deferred into '21. Some of it, it's a little early to tell. As I have touched on before, the extent to which this reprioritization sticks versus bounces back to what it was, it's just a little unclear. But some of it is definitely deferred work. It's a similar -- for a similar reason. We've got the [ POs ] in hand. We're ready to go. One project in particular in November -- late November, but the customer said, look, we've got the team that needs to starter is working over here, and they won't be back until January or February or something of that nature. So there's definitely some deferral, but there's sort of other stuff, it's a little early to tell. In terms of margins next year, we should see some margin expansion to Siraj's question before about the resource base and the operating leverage, we should see some of that coming through next year. We continue to invest in technology that accelerates what we do, and we take that to the bottom line as quickly as we can. So we should see some expansion, but it's a little early to tell as to what those numbers will be. And clearly, that will be part of the dialogue when we talk full year results in February.

Our next telephone question is from Tony Mitchell from Ord Minnett.

Look, just one question on -- well, I've got a couple of questions. But the first one is when you say that the second half is EBITDA is going to grow at 30% over the first half, maybe I'm missing something, but in the first half, you did $62.5 million underlying EBITDA, which implies that the second half is $45.5 million to $48.5 million, which is about a 15% drop. So can you explain why you're saying the EBITDA is going to go up 30% over the first half, please?

Yes. The number we have for H1 EBITDA, underlying EBITDA, is $49.1 million. And so we've done all our numbers on that. I'm not sure where the $62 million number that you referenced comes from. But essentially, H1 underlying $49.1 million and then we take our H2 range, apply the H1 exchange rates to that...

Right. Right. Okay. Okay. All right. Fine. The second thing is, when you say that 30% of your revenue comes in the fourth quarter, how does that apply to EBITDA? Is it a similar number for EBITDA being 30% for the full year?

No. It tends to be more, Tony, because we get some sort of operating leverage through the year. I don't have a figure off the top of my head, but it does tend to be more.

Okay. And just how big is -- are your Californian clients as a percentage of your business?

So we report that our top 5 clients are currently, Kevin?

About 88%.

88%. And they include the biggest tech companies in the world, some of which are headquartered in California, others on other locations on the West Coast, like Seattle.

And our next telephone question is from Michael Aspinall from Jefferies.

Okay. Just two quick follow-ups from me. You mentioned a couple of times that the engineers being put on other projects is one of the reasons that the revenue slowed down a little bit. And you also mentioned that Facebook had plans to hire 10,000 engineers later in the year, would you expect that hiring to have a direct impact on kind of the volume of data that they might need for you?

So the reference to the Facebook hiring, that was an article that appeared in April, more or less at the outset of the pandemic. You can find it online. So they're onboarding those people or they have been through the year and still are, one assumes, right? And yet, you would think that, that should be positive for the work that we do. But again, what we see on the ground is that the reallocation that we've -- or the reprioritization that we've been talking about. So our view of the workings of all of our customers is not complete. But certainly, across our customers, there's an intensity in their new product development, and that's impacting their resource management and that's impacting the work that we do. So I hope that helps.

Okay. Yes, that's good. And then just a final one for me. I mean you mentioned that 2021 should grow in line with industry growth rates and kind of the paragraph before that, you mentioned that the industry is expected to grow at 28% annually. Is that the kind of order of magnitude that we should expect in FY '21?

That's a fair deduction, yes.

Our next telephone question is from Ross Barrows from Wilsons.

Just a question on -- I missed the first couple of minutes of the call, so apologies if you repeat it. But just in terms of the EBITDA margin guidance that you have for the full year, is that still in place? Or can you make any comments around the profitability of the revenue that you are getting in the second half?

We're still working that through, Ross.

Yes. But I think the main point, Ross, we don't actually have EBITDA margin guidance. Obviously, we've just got the EBITDA -- underlying EBITDA range. But I think just think about the comments Mark just made around what we expect to see going to '21 around some kind of operational leverage and scale into margins.

Yes, sure. No, I heard the '21 comments. That's fine. And just for the second thing, you talked about the new business areas in shipping, automotive, education, et cetera. Just given the dominance of the top 5 customers that you have and also the observation the government is a big opportunity about emerging, I guess, which of those sectors is likely to be, I guess, the strongest for you? And over what time frame, difficult to say, but which is the most likely to be the most material of those 4?

The ones where we tend to get the most new business is if there is some sort of digitization effort in the particular use case. So automotive is very rich in its use cases and data requirements. We're even doing projects with model cars, model robot cars. And the customer is programming them, and we're collecting data on how they move around the course and stuff like that. So there's some pretty rich use cases in the automotive sector. The other ones that are mentioned there, I mean, there are all things that the customer wants to digitize a part of their operation, it could be routing with logistics and shipping, for example, it could be chat bots for communication and customer service in education, for example. So the richness in what we do is it can apply sort of across industries on a sort of an AI thematic like a chat bot, for example. So all that is to say that I can't kind of pick a winner at this point. I think automotive will be strong. E-commerce will continue to be strong. We don't mention that in that paragraph, but it's also another strong area. And then financial services, perhaps another one, but we continue to try to win as many customers as we can across as many use cases as we can for obvious reasons.

Yes. Great. Just the last one on your secure facilities. Can you just make any comments around that, whether they've been able to continue to operate at the levels that you would have -- that you were hoping for? Given the environment, obviously, if the projects are not there, then it would slow, but just in terms of what you do have, have there been any operational or execution issues given the environment we're currently in? Or is it actually been able to perform quite well?

So the highly secure stuff has performed quite well. We sought and were provided with exemptions given the nature of the work that we do in the highly secure area. And we've set up our facilities, all of our facilities, for COVID-safe practices in terms of screens and distancing and the like. The commercially secure work, we got exemptions from customers to do that in an at-home mode, and we are gradually moving that back into facilities as government guidelines allow, again, moving them back in, in COVID-safe practices. So overall, the level of work that we're doing and the quality of the execution has been strong or has been -- the work that we've got to deliver, we've delivered well. What's been impacted is just winning new customers, and that's independent of whether it's secure or not, just the sales motion, the marketing motion, everything that we've traditionally done face-to-face. It's fine virtually from time to time, but it's not -- it loses currency if it has to be sustained over many, many months. So the issue is less whether centers are operable, more winning new work to put into them.

And our next telephone question is from Lucy Huang from Bank of America.

So I've got two follow-ups. So firstly, just wondering in terms of your sales efforts moving forward, how should we be thinking about which sectors or type of work it will be geared towards? And I guess as part of that, what's your long-term thinking about the relevant work from tech giants? Do you think that will still be a major driver of growth in the industry over the next few years? Or do you think it's really the new product development industry sectors that will be the key drivers moving forward? And then just my second question, just wondering with your top 5 customers, what's the rough with split of revenues between each of them? So I understand you can't reveal their names, but just wondering whether it's quite diversified across supply? Or do you have some customers where you are generating a lot more revenue out of?

Yes. So to the second one, first, Lucy, we don't disclose the split amongst our top 5 customers, but yes, there's more in some and less in others. To the first question around Relevance. Relevance will continue to be a strong and growing part of our business. We don't want to send the impression that our customers are forgetting about Relevance. They're not by any measure. What's been behind the result is a reprioritization away from some projects, some of which are relevant. So the customer has taken a view that I've got so many resources, the new product development is very important, where can I shave some resource with manageable impact in order to get accelerated development of these new products and they've made their decisions accordingly. So we don't want to give the impression at all that Relevance is going to continue to be a strong part of our business and continue to deliver high growth, so I'm sorry if we get that impression.

Our next telephone question is from [ Andrei ] from Affinity Asset Management.

Mark, could you just help me wrap my head around what kinds of projects are actually getting kind of postponed? Like, obviously, you cannot name specific clients, but just in very simple terms, is it basically saying, okay, I will defer updating my search engine algorithms or my ad-serving algorithms and instead, I would put this data scientists on developing like e-commerce recommendation? Or like what kind of use cases are so important? And what kind of use cases are becoming less important out of the pandemic?

Yes. Thanks, Andrei. Look, we can't kind of go into the sort of specific detail there, but perhaps you could look at it this way. If you have a product that has dominant share of the market, you could probably be a little more forgiving with your investment in that product vis-à-vis one where you're trying to grow and enter and win share. So when we look at the projects that have been impacted by this resource allocation, they tend to be the products of that nature, where the customer has a really strong grip on the market so they can tolerate a bit of reallocation. As we've discussed in a number of the responses, the extent to which this sustains is -- it's too early to tell. And I think over time -- and we've seen this before. For example, with competing search engine companies, if one were to take their -- relax a little bit on Relevance data, the other one starts to gain share. So I doubt that it's a long-term permanent shift, but the degree of bounce back and timing is just a little unclear because we're not around the product management table with our customers, unfortunately. Does that help?

Yes. It definitely does help. And so then a follow-up from that would be, is there a natural cycle to this project? They have reallocated the data scientists so to the new projects, is there a natural cycle where generally data scientists spend, whatever, 1 to 3 months kind of scoping out the projects, thinking what kind of data they actually need? And then it come -- they come to you for data? Or like is it more equally distributed across the project?

So it varies depending upon the product and the use case, Andrei. The refresh rate for Relevance tends to be fairly high. So you've got to keep feeding data into those algorithms. A language-based product can require less frequent, but still some refresh as they want to add new languages or the -- just like the speech recognition program to say it doesn't work under certain conditions. I don't know, outside when it's windy, they need to collect more data. But with a lot of use cases, there is a surge upfront that requires a lot of data and a lot of work and once the product starts to work as designed, they have a better grip on what sort of data they need in it. But overall, pretty much all AI products needs some data management and data maintenance, models drift over time. They start to do the wrong thing. And so you need to sort of course correct them to make sure that they're behaving and doing the things that you need them to do. So yes, there's no clear answer to your question, unfortunately, it does depend upon the use case. And I think the general principles of model drift and data update are good for our business for obvious reasons.

Yes, for sure. So my question, I guess, was more on that point, you touched upon on the lag before the surge upfront happens rather than on the refresh rate as such. Like, does it -- is there kind of a typical time -- a typical lag you would expect from data scientists to kind of to develop the scope requirements to develop their data needs before this -- from kind of the data scientists being allocated to some upfront search for the initial development work for data needs?

Yes. So just a real subtle distinction. So we're not talking data scientists and design here, we're talking engineers and using data. You're correct in saying that a data scientist may have a more intense phase at the beginning of the product at the design phase. But the engineers are more in the build phase and the maintenance phase. And the intensity of their work, as I said, depends upon the use case, depends upon the nature of the data, it depends upon the data refresh rate, et cetera. So there's no clear answer to your question, unfortunately. But if I could come at it another way, I guess you're essentially asking, well, they move these resources from here to there, when are they going to move them back? As I said, we don't quite know. But the principles of model drift and data update mean that at some point, they've got to reattend to those original products.

Yes. That's very interesting. Maybe just last one for me then. These engineers, you mentioned it's more relocated engineers than data scientists. So is it data engineers? Or is it kind of your broader software engineers? And if it's just software engineers then kind of shouldn't the data requirements largely flow from data scientists to data engineers? And why does it reallocate your -- the data requirements so much if it's just software engineers?

Andrei, I don't know the answer to that question. But what we know is that our major customers have reprioritized some of their engineers, not scientists, engineers, whether they're data engineers, software engineers, design engineers, I'm not 100% sure, but they've reprioritized engineers towards forward-looking new product development away from more mature data-hungry products, and that's impacted our revenue. The exact nuance of what's going on underneath the covers in all my customers is just something I don't have that clarity to. I'm sorry.

Yes. I think the important takeaway, Andrei, is that engineers are needed to load the data, right? If you don't have the engineers, you can't load the data. If you don't -- they don't load the data, we can't do the work. I think that's the takeaway.

So I think at this point, we may have to wait 10 minutes to the top of the hour. We can take questions from a couple more.

Excellent. And we have two more questions for today. Our second last question queue is from Garry Sherriff from RBC.

One quick follow-up. Given the search advertising revenue was so strong in the recent quarterlies from Facebook, Google, Microsoft, I mean, is there any risk that the Relevance work you're doing on those major projects are being internalized? I mean, I know Google are doing more of the data sourcing aggregation themselves in-house. And I just wonder, is this perhaps a permanent shift in terms of them vertically integrating downstream into some of that stuff that you currently do? Any thoughts would be greatly welcomed.

Garry, we don't see that. We'll reiterate the same response. It's a reprioritization to put resources towards new product development, we don't see any work coming in-house.

And our final question today is from Siraj Ahmed from Citigroup.

Mark or Kevin, just a quick -- just to think, you've given the guidance for the full year, can you just help us with what the revenue growth rates or the margins are? It sounds -- my estimate is revenue growth rates for the second half is going to be around 14% to 15% year-on-year, is that fair?

We haven't disclosed revenue growth at this point, Siraj, sorry.

There are no more further questions in the queue, I'd like to hand the call back to today's presenters. Please continue.

Yes. Thank you, and thanks, everybody, for joining the call. Thanks for your questions. Just as a summary, and I think we've made the point throughout the Q&A, the major factor here is this reprioritization of resources. And I'm sorry, I don't have all the detail requested around exactly what they're doing. But fundamentally, the projects we work on need engineers. And if the customer decides to reprioritize them, then the project sees less work from us and that goes to less revenue. We think the positives in all of this are that we're involved in, as I said, the number of new projects amongst our major clients has increased by 32% year-on-year. And that gives us many more points of presence inside of our customers. And as we've seen historically, some of those projects can grow to be quite large for us, which is very positive. So we have a near-term impact due to this issue. But as I said, also, the pipeline for next year is looking pretty positive. And we're looking forward to talking to talking you all at the end of February for our full year results and taking you through a view of 2021. So thanks once again for joining the call. I hope you have a good holiday season. I hope you stay safe from the plague. And I look forward to seeing you all in results season. Thank you very much.

Ladies and gentlemen, that does conclude the call for today. Thank you for all participating. You may all disconnect. Goodbye.