Yelp Inc. (YELP) Earnings Call Transcript & Summary

So AI, a topic that, I think, we're having with every company. Let's -- I want to take it on the positive side and perhaps the risk side. Starting on the positive side, how are you using AI today? What are some of the ways you think Yelp can benefit given the review first-party data that you have?

Yes. So there's obviously been a tremendous amount of conversation about large language models over the past 3 months. In fact, maybe it's the only conversation around AI that we've been having recently, but it's certainly not the only way that people are applying AI, and we've been using AI for some time in the form of machine learning, whether that's for search or for these matching algorithms between consumers and advertisers. So we're very experienced using these types of methods. Specifically to large language models, we did start using them last year, so it's not new to us. And we've applied them in a couple of ways that we're already delivering value. One of them is we announced as part of this batch of products, more than a dozen that we announced a couple of weeks ago, which is people just want to see quick snippets of reviews to really understand what people are saying about that particular business. And large language models are excellent at really summarizing and extracting sentiment and information from large data sets, obviously. So this has been a high-value add for our visitors. Another less obvious one is, in search, it turns out, I didn't quite realize this myself until recently that synonyms really matter because people search for things, inquiries in different ways for the same thing. And so large language models are very, very helpful in identifying and prioritizing which synonyms to use for search. So those are a couple of things that we've talked about that are important. But I think more broadly, the opportunity is really to make something like Request-a-Quote more conversational. Today, it's a tree that you answer. And as you give an answer, it may take you down a certain path in order to gather that information. I think there's a lot of potential to make that whole interaction more conversational. On the Service Pro side, a lot of times the Service Pro might just reply call me. They could be a little more expansive in what they're sharing with the consumer. And so there's an opportunity, we think, to perhaps harness large language models. So broadly, we see this as an opportunity, especially since we have very reliable information for consumers between the ratings, but more importantly, through the reviews. And we think that, that is a real asset for us. We've talked -- we've heard a lot about the way that these large language models hallucinate, it does seem that it's inherent to this particular way of constructing a sense or a paragraph. And so the ability to train a large language model on trusted content, we believe is a very powerful way to enhance the reliability and trustworthiness of what that large language model is yielding. So overall, we think between our content, the infrastructure we've built, our ability to deliver that to consumers, we overall see this as a real opportunity for us.

So on the risk side, I mean, how do you protect those reviews? So certainly, a lot of concern around online publishers around just the data being scrapped and then instead of going to the website and getting ads you're seeing a summary in Bard or in ChatGPT. So I just want to hear about how you plan on -- or how you can protect your review data? And how big of a risk you think is to Yelp and how you can address it?

Yes. And just to point out the obvious, Google already does a lot at the top of the page and on the second page and on the third page, in some instances the fourth page. And so organic results are already pushed very far down. So this is a new. And we've been competing with Google for more than a decade quite successfully. And in terms of summarizing reviews, there's also a lot of summary already occurring on Google, whether it's the answer box that they use or even in how the search algorithm has prompted publishers to increase the amount of information that's showing up in the organic result itself. And people are still clicking through. Why do they still click through because they want to go to the place that is the origin because that is where you have more information, where you get more of the insights into that decision making. So I would just say we don't see large language models in and of themselves as transforming that dynamic with Google itself. Of course, summarizing our results for consumers and obviating their need to come and visit us. So I would just say that's the first part of the answer. The second part of the answer is, okay, well, that's all well and fine, but what about fake reviews? How much of a threat could that be? Now Yelp is going to have his 20th anniversary next year. And we spent 19 years figuring out how to discern fake reviews from authentic reviews, and it's not easy. And we believe we have a lot of techniques in place that enable us to attack problems like this. So of course, we have to figure out exactly how it's going to work with these large language models. I think the large players have committed to watermarking or otherwise creating ways to discern when a response is generated in a large language model. So that certainly will be helpful. So they're, I think, going to play their part. But also there are techniques that can be applied ironically using AI to discern when a response has been generated by a large language model. So we think that the ability to moderate content is actually a real asset at Yelp. In fact, we don't even display approximately 25% of the reviews we receive to ensure the reliability of those reviews in the first place. So we're very, very committed to ensuring that reliability and what does that translate into, we believe it really translates into consumer trust, which I've talked about as being so important overall to the experience on Yelp.

So following up, you mentioned traffic in Google. I think it's always been a topic for Yelp over the years. I also think you've significantly reduced your reliance on Google traffic over time. But any update on how much of your traffic today is direct to the Yelp website, through the app, through Google or other means? What are you willing to share there?

So what we've shared is that the majority of clicks on Yelp do come through the app. We obviously attend to it and it's been more than a decade that we've really focused on the strategy of enticing, encouraging people to use the app as opposed to using the website. And we think we've been successful with that. And the beauty of the app, of course, is that it provides a very rich experience. And I think like others, we really do want to encourage folks to use the app in order to ensure that they're getting the best possible experience and it also limits our exposure to Google. That being said, of course, Google generates traffic for us. And so what we want to do, of course, is optimize the way that we present that information to Google, but probably even more importantly, what we want to make sure is that when we do get that click through, that we're really capturing the opportunity. I mentioned that we're investing heavily in the consumer experience. Part of that is to modernize it. Part of it is to make it more visual. But part of it is to really ensure that, that landing page experience or when you deep link into the app, that there's continuity in the experience that what you see is relevant to you and then you choose to continue to use that app. So that engagement is very important to us. And it's also very important because we want people to write reviews. So we want them, when they come to write that review, obviously, that more engaged user comes back more frequently. They might try services if they've only tried restaurants. They might try restaurants if they only trade services. So it's really a whole effort -- a broad effort in order to ensure that engagement and to maximize the value of the traffic that we do receive from Google.

So you've mentioned a few times now you've been working on improving the consumer experience for the past year or so. I know you don't disclose traffic metrics quarterly anymore, but could you broadly talk about consumer engagement trends that you're seeing? And are the initiatives that you're working on working to get traffic back to where it was pre-COVID?

Yes. So maybe just as a way to make it tangible, in our Request-a-Quote flow, actually, the number of requests declined in the first quarter by 5% from the first quarter of 2022. And yet, clicks were up. Services had record revenue. So how can those two things be true? And the way that they're true is that we are able to drive that engagement with the consumer, understand their intent better and better match them to the business. And so we feel like we've made good progress around engagement. First quarter, we did see good progress in encouraging people to write reviews. That matters again, both because it's the purpose of Yelp, it's our content that we're monetizing, but it also deepens the relationship with that consumer. So I think that's a good example of how clicks can differ, for instance, from MUVs or Request-a-Quote requests and still enable us to be successful in an environment where with inflation, consumers are doing less things less frequently. They're spending about the same, but they're doing less frequently. So we have to be able to manage that. And all of that, again, as a result of that product-led strategy that we set for ourself.

Let's stick with CPC and ad clicks. That is something you do provide. Most of the growth has been on the CPC side recently. A lot of that due to funky comps, but ad clicks did turn positive in 1Q. What should we expect from here? I mean, how important is the return to ad click growth?

So just as a reminder, when advertisers advertise on Yelp, they provide us a budget, and we optimize that budget on their behalf. We optimize that budget by running an auction. And we believe that auction finds the market clearing price for that visitor at that time, in that geography for that category. So obviously, if we're generating robust advertiser demand and visitors are about the same, there's a market clearing price that's going to have to rise. So CPCs did go up, which is why this lead through rate becomes all the more important because if I'm paying more for a click, then I expect more value out of the lead that I receive. So we are not choosing or setting the click or the CPC. What we're doing is we're optimizing the deployment of that budget on behalf of the advertiser, and we really care because if they don't perceive that they're receiving value, they will churn. We don't want them to churn. We acquired them, we want them to be successful on the site, and we obviously want them to expand their spend. So we don't think of it in terms of, hey, do we want clicks to be higher? Or do we want CPCs to be higher? We think of it as we're going to invest to continue to make the matching process better and better to deploy as efficiently and optimally on behalf of the advertiser to deliver them long-term value.

So shifting to a few questions on financials now. You raised your 2023 revenue outlook after 1Q. Can you walk us through some of the macro assumptions you're making within the guidance? And then one question we get a lot is just how much visibility do you have given the month-to-month nature of your contracts?

Yes. So I would just say that I think the first quarter was the 13th quarter in a row that we were ahead of analyst estimates in terms of revenue. I think that suggests that we have some experience and visibility and understanding of the dynamics on Yelp. So that's been something that we've built over time. And what I would remind folks about is that Yelp is a diversified set of categories. I think that diversification at scale is really important to being able to forecast because what we saw during the course of COVID was that Restaurants, Retail & Other revenue did decline, but Services expanded. And so we see that even within each of those broad category sets that there are dynamics and shifts way -- in the way that people are spending. So the first thing I would just say is diversified side of categories is very important for being able to forecast. The second is we do use a lot of scenario planning in order to play out what different paths could be. And then we do develop a playbook and think ahead of time about how we might react under different scenarios so that we can be proactive and not reactive. And then, of course, the guidance we give has to reflect those risks and uncertainties. And I think we'd all agree, it's hard to know where the Fed terminal rate is going to be, and it's very difficult to forecast when if we're going to have a recession and what the depth of that recession might be. So I think we're operating right now under conditions that are somewhat heightened in terms of that macro uncertainty. And so we obviously take all that into account. All that being said, we are very focused on delivering value to advertisers, doing it more efficiently, monetizing the services leads that we have, creating a great consumer experience. We're focused on the things that are in our control. And again, I would just offer 13% growth in the first quarter, I think, stands out among ad peers. And so I think it's a reflection of the performance that we've been able to deliver. And then as we go through the year, there's always some seasonality historically that we've seen, where revenue does increase as we move quarter-by-quarter through the year.

On the expense side, you guided margins to pretty meaningfully increase in the second half of the year as well. What's driving this inflection? And could you just help bridge where you expect the OpEx decreases to come from as we work through the year?

Yes. So there's a few dynamics at play in the first quarter. One most certainly is payroll tax that doesn't recur. Another is seasonally ad spend declines in the fourth quarter. It increases in the first quarter, and then it will decline again, we'd expect in the fourth quarter. So that's part of the expense profile. Obviously, I mentioned that we had higher headcount than anticipated in the first quarter because of retaining employees, again, overall positive. But we do plan to manage to headcount being about flat by the end of the year. So that's also an important element. So just in terms of margin, with the expectation, obviously, our Q2 guide at 3 20 to 3 30 is up from the 3 12 that we delivered in the first quarter. And then the overall guide for the year being up from where we initially guided. We expect that revenue to increase as you move quarter-by-quarter. And then on the expense side, we think we can manage that so that you basically get that expansion in margin as you go through the year.

How do you think longer term just about the right level of growth, sustainable or healthy level of growth for Yelp? And has your view changed at all given your performance over the last year or so?

So we're very, very disciplined from an ROI perspective, and we really think and put a lot of effort into determining, "Hey, is that next sales headcount going to yield? Will that next dollar spend on marketing yield? Will the next engineer yield?" So first and foremost, we're always looking at investing dollars efficiently to drive performance. But we do -- we are committed to delivering profitable growth over the long term. We think that there's a balance between those two that enables you to sustainably and consistently deliver performance over the long term. We think that this product-led strategy gives us an opportunity to drive real margin leverage. I'd just remind folks, we made the transition on the sales headcount. We said in '21 and '22 that those were going to be investment years for product and engineering. This year, we committed refined to have flat head count this year and yet revenue is growing from the 11 94 that we did in 2022 to about $1.3 billion this year with that anticipated flat headcount. So we think that we are on that point where we can deliver the margin, but we do think in terms of trade-offs between margin and growth.

Capital allocation. You bought back a ton of shares over the years. You've been recently buying about 50 million a quarter. Why is this the right buyback level? And what would have to change for your priorities to shift on the capital allocation front, say, M&A, a dividend or something different?

So just to remind folks on the capital allocation front, we are committed to returning capital in excess of a target balance to shareholders. The way that we arrive at that target balance is there's an operating amount of cash that we want to have, there is a buffer to that. We do hold cash on the balance sheet for tuck-in acquisitions. And then beyond that, we want to return it. When we think about returning capital and the level that we purchase at, we do use a pricing grid, as you would expect, to have a specific level of purchases. And as you'll note, we have been able to reduce share count over time, and we've returned approximately $1.2 billion in capital to investors. Now what would cause us to shift just say we're very disciplined on the M&A side. We're unwilling to overpay for assets. From a tuck-in acquisition perspective, the things that we would be interested in doing is continuing to enhance the ad tech stack that we've built for the consumer, perhaps enabling them with more tools as we add things like including videos in their reviews, perhaps there's something that we can do around that. For Service Pros, we want to make life easier for them. Maybe there are things that we can enhance for that experience. So those are the examples of the kinds of things that we would use to deploy capital that differently than just the share repurchases that we've been doing.

So I'll end on the exciting topic of stock-based comp. You've laid out a plan to get stock-based comp down from 13% today to less than 8%. And by the end of 2025, help us with how you get there? And then a question we get a lot too is how much of this is going to be offset by higher wages?

Yes. So we -- the way we think about it is both revenue and expense. And we have given ourselves the target of getting stock-based comp as a percentage of revenue to 8% by the end of 2025. Obviously, we are very focused on continuing to drive growth. And then in terms of expense, we are going through a process to look at how we can shift cost from equity to cash and the way that we compensate people and continue to motivate them. And obviously, notwithstanding the [indiscernible] environment, it's going to be continue to be very competitive for product and engineering talent and good people. So we recognize that and so that is complicated. And we're right now in the middle of thinking through all of that for 2024 and for 2025, but we do see a path, obviously, to being able to deliver that by the end of 2025.

Awesome. Well, thank you very much.

Thanks so much.