|Shares Out. (in M):||65,816||P/E||20.3||16.6|
|Market Cap (in M):||2,632||P/FCF||14.8||12.0|
|Net Debt (in M):||-348||EBIT||143||211|
What Does Criteo Do?
If you have ever looked at a ticket on Expedia and then noticed advertisements on other websites for similar flights, you have been the subject of Criteo retargeting. Criteo provides display (image based as opposed to video) retargeting for more than 12,000 companies including Walmart, Best Buy, Macy’s Dick’s, Sears, eBay, Microsoft, Samsung, Visa, Intuit, Adidas , Airbnb, BMW, Expedia, Trivago, and Travelocity. No customer accounts for more than 2% of sales. 66% of revenue is from retail verticals, 16% from travel, and 18% is from classifieds (customers like eBay), autos, and telecoms. Geographic revenue mix is 39% Americas, 38% EMEA, and 23% APAC. In order to generate effective retargeting advertisements companies need access to their customers’ websites/apps to assess customer intent (demand side) and access to publisher inventory to display ads on platforms likely to attract customers (supply side). Criteo is well integrated on both sides of the ecosystem having observed over $500 billion in sales on their customer’s websites/apps (more than Amazon and eBay combined) and generating over 8 billion advertisement clicks on the websites of their 14,000 publisher partners over the past year. The process of (1) analyzing customer intent across devices, (2) determining the best inventory placement and advertisement display, and (3) assessing how much to bid for inventory based off different ROI targets, all in milliseconds, is technologically complex and requires advanced machine learning algorithms to be done effectively. The Criteo Engine that drives this algorithmic process is considered the best in the world.
Criteo’s revenue model is unique among ad-tech companies in that they only receive revenue on a CPC, or cost per click, basis, meaning that they only get paid when their advertisement leads to a customer click. This is in contrast to the more popular CPM, or cost per impression, basis, in which customers pay if an advertisement appears on a site, even if it is not visible.
Criteo benefits from secular growth in Ecommerce
Criteo’s sales are highly levered to the growth of Ecommerce. According to the Census Bureau, Ecommerce has grown at a ≈15% CAGR over the past 5 years in the US, and eMarketer forecasts Ecommerce to grow ≈21% per year worldwide through 2020. Weighting eMarketer’s regional forecasts by Criteo’s current regional revenue share, Ecommerce sales should grow between 15% and 18% in Criteo’s current markets through 2020 (although CRTO is growing their revenue share in the faster growing APAC markets). It seems reasonable to assume CRTO sales grow at a similar 15%-20% CAGR over the same period, before taking into account changes in market share.
Criteo is addressing a small fraction of its TAM/Customers
Up to this point, CRTO has been focusing solely on traditional display advertising. They just announced their intention to enter search retargeting (display ads on Google) which they estimate to be a $6 billion revenue ex-TAC market, and have yet to enter acquisition marketing (advertising to attract new customers), which they estimate to be a $9 billion revenue ex-TAC market. These two markets increase CRTO’s TAM from $10 billion to $25 billion and leverage CRTO’s existing technology and relationships. In addition, although the technology isn’t ready for video and offline retargeting, these two markets offer large revenue opportunities for CRTO.
Moreover, CRTO has primarily served retailers while getting little business directly from brand manufacturers (P&G, Kellog’s etc.). The recent acquisition of Hooklogic not only gives CRTO access to new inventory on retailers’ sponsored product pages and the opportunity to improve their performance through CRTO’s data and algorithms, but it also gives CRTO the opportunity to cross sell Hooklogic’s manufacturer customer base . Finally, CRTO is just entering China, the largest Ecommerce market in the world, in a meaningful way in 2016-2017.
Criteo has sustainable competitive advantages
Only Google sees more transaction data than CRTO. More data leads to smarter algorithms, better retargeting, and higher click rates. Higher click rates enables CRTO to pay more for inventory than competitors and increases the supply with which they can distribute the best ads. This improves client ROI, which attracts more clients and increases the amount of data. I believe this is a sustainable positive feedback loop.
As a result of their scale and focus, CRTO has the best results in retargeting. Here’s some evidence showing the superiority of their product and the quality of their economic model:
Quality Economic Model
Although Google and Facebook retargeting is not yet as good as Criteo’s it is certainly possible that their technology improves over time given their resources, scale, and talent. However, even if their technology becomes equivalent to Criteo’s, advertisers are still likely to prefer CRTO for three reasons. First, many of Criteo’s customers compete with Google and Facebook and feel uncomfortable giving them access to proprietary data. Second, while Google and Facebook exclude each other from accessing the competitor’s inventory, Criteo has access to both Google and Facebook inventory. Third, Google and Facebook’s cross device tracking is limited to their own platforms and the data is less transparent for customers. Because of these factors and the fact that it is inefficient to have multiple retargeting providers, Criteo can continue to gain market share even if their technology is only equivalent to these competitors.
Why Does the Opportunity Exist?
Header bidding is a new advertising technology that gained popularity in the US in 2015. Essentially the technology changes the auction dynamics for inventory. Instead of having a waterfall auction with a limited number of bidders at each level, CRTO will have to bid against more competitors. Moreover, instead of being a second price auction in which the party that puts in the highest bid ultimately pays the price of the second highest bid, header bidding moves to a first price auction where the winner actually pays the price that they bid. Ultimately, header bidding has benefited publishers and hurt buyers through higher CPMs for inventory. In response to higher inventory cost, CRTO can either (1) buy fewer impressions, which lowers CRTO’s gross revenue, (2) buy inventory at higher prices without passing on the price increase, which lowers CRTO’s take rate, or (3) pass the price increases to advertisers through lower ROIs. Prior to 3Q I thought the effects of header bidding were evident in declining take rates in the US. However, US take rates increased in Q3 so I’m less convinced of this now. It’s possible that CRTO shifted strategy from lowering its take rate to buying fewer impressions, which would explain the less impressive gross revenue growth this quarter. Nevertheless, despite the potential effects of header bidding, CRTO has been able to beat revenue and EBITDA targets every quarter since header bidding became popular in 2015. Over the same period the stock declined from $57 to under $25 as the TTM P/E multiple compressed from 43x to just 20x today.
CRTO argues that header bidding is a “low single-digits type of issue.” They believe that buyers will start lowering their bids over time because there hasn’t been a fundamental change in supply and demand that should change the value of inventory. They also believe that their superior technology will eventually enable them to predict the second best price in the new auction format so that they can bid the minimum amount above that to win the auction. As a result, CRTO may eventually pay a lower price than before header bidding, while those with inferior technology will subsidize the more advanced players. While I agree that header bidding doesn’t change supply and demand, it does increase the ability of buyers to express their demand through a broader auction (more liquidity). Therefore, mild price increases for inventory seem rationale to me. Furthermore, given the higher CPMs for publishers, it is likely that header bidding expands to Europe and APAC over the next year or two. I also would not count on CRTO being able to predict other bidders’ pricing, given the difficulty of doing this in the stock market.
On the other hand, I think it is likely that advertisers will start accepting lower ROI targets as they become more dependent on Ecommerce for growth and shift their advertising budgets from other forms of advertising like mail and catalogues that are even lower ROI and less measurable than retargeting. Also, header bidding has a negative impact on Google, and given their power over the entire internet advertising ecosystem, they have been pushing back against header bidding and may slow its adoption or change the technology in their and CRTO’s favor over time. If these scenarios play out, header bidding would have essentially no impact on CRTO.
Ultimately, although I think header bidding is a minor headwind, I don’t think the issue comes close to justifying the stock decline (down 56% from the July 2015 high to the February 2016 lows) that came as the company continued to grow revenues 33% and EPS 89% YoY over the first three quarters of 2016. I incorporate aggressive header bidding effects in my base case by lowering CRTOs gross revenue growth rates in the sell side model I am using (which has average sell side expectations) by 200 bps each year through 2020 and reducing CRTO’s take rate from 40% to 39% (more than the 70 bps decline in take rate in the US since the end of 2014). Even with these effects, EPS would grow at a 17% CAGR through 2020.
Many investors are reluctant to buy any company that competes with Google and Facebook. This is probably a good rule most of the time, but I believe this is an exception for a few reasons. First, as explained above, I believe many advertisers will not use Google or Facebook for retargeting because they are uncomfortable giving them access to data and they don’t offer access to each others’ platforms. Second, Google and Facebook benefit from their relationships with CRTO because CRTO is one of the largest buyers of their high CPM inventory. I don’t think it would be in their interest to close off a high CPM bidder from their platforms (for reference Google accounts for about 20% of CRTO’s inventory spend and Facebook accounts for about 10%). Third, CRTO is a trusted independent partner that helps Google and Facebook develop their advertising technology for the wider ecosystem. For example, CRTO is often the first to test Google’s new advertising products, was the first to test Facebooks mobile advertising platform, and has a custom integration with Facebook Dynamic Product Ads. Fourth, Criteo is singularly focused on retargeting and at this point has better performance than Google and Facebook’s performance isn’t even comparable. These relationships require monitoring but I believe CRTO has found a strong niche in the ecosystem. An incremental positive for CRTO, would be integration with Snapchat, which they are not on yet, as it could reduce the power of these large suppliers.
Ad-Tech “Peers” Overhang
Ad-tech has been crushed over the past twelve months. FUEL is down 49%, RUBI is down 53%, TUBE was down 40% before being acquired, while CRTO has managed to fall just 11% (CRTO’s decline came earlier and they are down 36% from the recent $57 high). Despite having very different business models than these public competitors and CRTO having been able to overcome some of the issues that caused their poor results, I think the sector’s performance has contributing to the downturn in CRTO’s shares.
FUEL ($459mm market cap) was the closest public competitor to CRTO, as they offered a retargeting service, but they have shifted from managing campaigns to offering their technology as a service to ad agency trading desks after a string of revenue misses and guide downs. This shift was likely due to their inferior technology and CPM revenue model when compared to CRTO. In the most recent quarter they gave guidance 18% below the street and announced their new CFO was leaving the company after less than a year on the job. With no earnings, increasing leverage, and a $100mm market cap, they simply don’t have the scale to compete with CRTO. I think their losses are more likely a reflection of CRTO’s strength then the industry’s weakness and they will ultimately go bust or be acquired.
RUBI ($298mm market cap) is a SSP (sell side platform) so it is in a very different business, but their disappointment has been due to (1) header bidding (which is especially bad for SSPs because inventory can be sold before it ever reaches the SSP), (2) the acquisition of Chango, a retargeting competitor to Criteo that is not meeting expectations, and (3) the shift to mobile. In the last quarter, sales guidance was 11% below consensus, EBITDA guidance was 17% below consensus, and EPS guidance was 33% lower than consensus.
The above list excludes CRTO’s most direct competitors, which have been acquired. Twitter acquired TellApart for $535 million at an estimated 5x EV/S in April 2015. Tesco acquired Sociomantintic in April 2014 for around $200 million, an estimated 2x EV/S. Struq was acquired by Quantcast (private) in October 2014, Triggit was acquired by Gravity4 (private) in March 2015, and Dotomi was acquired by Conversant for $295 million in November 2013. This was a very crowded field and CRTO appears to be the winner in a winner-take-all market.
In contrast to their public competitors, in their most recent quarter CRTO ($2.6b market cap) beat sales estimates by 2%, EBITDA estimates by 19% and EPS estimates by 30%. Guidance for the fourth quarter year was also above expectations. Simply put, CRTO’s business model and performance has been completely different from its public competitors, yet I believe their performance has been dragging down CRTO.
Despite having grown revenue 33% and EPS 74% over the past year, and consensus expecting 18% annualized revenue growth and 22% annualized EPS growth over the next 3 years, CRTO is trading at just 20.3x trailing earnings and 16.5x forward earnings. Furthermore, consensus is expecting just 15% “core” net revenue growth over the next two years (core excludes revenue from search and the Hooklogic acquisitions), which is at the low end of the 15% to 20% Ecommerce growth expected through 2020. Given that CRTO has consistently taken share, this level of core revenue growth seems highly conservative. The market is also likely underestimating non-core revenue growth through search, Hooklogic, China expansion, and video opportunities.
To value CRTO I created three scenarios. The primary differences between the scenarios are core gross revenue growth rates and take rates. My base case assumes core revenue grows 16% per year through 2019, just 100 bps more than Ecommerce. I also lower the take rate from historical levels of between 40% and 41% to 39%. I put an 18x multiple on 2019 EPS of $3.15 (mainly because it gets me a similar value to what I get through a DCF), discount it back at 10%, and subtract cash to get a value of $48 (+20%). My bear case assumes core revenue grows at 13%, slower than Ecommerce, and take rates fall 200 bps. This results in 2019 EPS of $2.48. Putting a 15x multiple, subtracting the cash and discounting this back at 10% results in a value of $33 (-17%). Finally, my bull case assumes core revenue growth of 20% per year, at the high end of Ecommerce growth, as CRTO takes share and expands to new markets and verticals. This results in 2019 EPS of $4.06. Using a 20x multiple I get a value of $66 (+66%).
It’s also worth noting that TUBE, an unprofitable $540mm market cap ad-tech company which focuses on a cross platform, video-based SAAS product was recently acquired by Adobe at 2x EV/Sales, 3x EV/Gross Profit, and 41x EBITDA (all multiples based off of 2017 estimates). IF CRTO were to trade at these multiples, it would be valued at $71, $46, and $164 respectively. I am not arguing that CRTO is worth $164, but these multiples give me more confidence that the downside is limited.
Overall, while the risk reward is less favorable than before 3Q earnings, I still think this is a rare opportunity to buy an industry leader riding secular growth with 65% upside and a favorable risk/reward.
|Entry||11/18/2016 08:03 PM|
Thank you for the write up. How does the shift to mobile affect their business model? In other words, as users shift transactions to specific apps, does that box them out altogether. I know retargeting does exist in mobile, I just think the game is different when I go to my Amazon app to buy X and my Expedia app to buy Y. Once I'm in a given app, I'm much more captive and commited and it seems a business like Crit is much less effective.
|Subject||Re: SteelHouse Lawsuit|
|Entry||11/21/2016 12:29 PM|
Regarding the Steelhouse lawsuit
-Both sides dropped their claims on November 1
-It’s important to note that SteelHouse was countersuing the claim that they were stealing click attribution from Criteo, and five former SteelHouse clients (now with Criteo) backed Criteo’s allegations. If I were Steelhouse, which is already losing the retargeting battle to Criteo due to lack of scale, I would also respond aggressively to Criteo's claims because they represent an existential threat to their business.
-Criteo's does not have long-term contracts with clients. If clients don't see an increase in sales (which they measure internally) they can leave Criteo at any time. If such a large proportion of the clicks Criteo says it generates were from bots, which don't generate sales, clients would reduce spending and leave Criteo. Instead, same client sales increased 15% last quarter. To me this part of the countersuit makes no sense:
impressions. If the client is unhappy with the number of clicks and decides to increase the CPC from 50 cents to $1.00, Criteo could continue to bid on the exchange at $2.00 per thousand impressions, but now gets 50 cents more per click, increasing its profits while not actually increasing customer performance. This model deceives customers. Criteo’s customers believe that an increased CPC bid will yield better performance. But in reality, the increased CPC bid does not result in any additional traffic to the advertiser, because the exchanges occur on a CPM basis.
Based on how I understand Criteo’s business this is just false. Clients authorize CPC levels based off of internal ROI targets. If a client authorizes Criteo to increase its CPC, all else being equal, it is effectively agreeing to lower its ROI target in order to increase volume. If the lower ROI target does not yield additional volume, the client will see Criteo as a terrible investment and stop spending on the service. They are certainly not going to increase their spending without any benefit to themselves, which they can measure internally through A/B tests.
-Perhaps 55% of clicks could not be attributed by SteelHouse simply because they don't have the scale/technology to attribute clicks on mobile and across devices, which is technologically difficult but an increasingly large proportion of eccomerce sales. Criteo is pushing hard to develop its cross device graph precisely so that (1) clients attribute more sales to CRTO and (2) competitors can't take credit for sales that should go to Criteo, which was Criteo's accusation against Steelhouse in the original lawsuit. Here’s what the CEO said about the cross-device graph in the last investor day:
In 2017, we plan to offer the user graph as-a-service. What this means is, as an advertiser, as you pass to us your ID, we will pass you back, at a granular user level, the consolidated Criteo user ID that were built from all the interactions of our network. We will pass this back to you so you can use it in your own tools for attribution, for measurement, perhaps for other purposes too like site personalization and given that key, if you like, to the user ID identity, that stretches across all their users, not just the ones working with Criteo. We will pass them the full Criteo-device graph. We will offer this as a free service to advertisers, but there are three main benefits for Criteo from doing this and we will go through each of these three quickly. We are doing this in order to increase sales that are attributed to Criteo. We are doing this in order to improve the scale of our cross-device graph and we are doing it to deepen our relationship with advertisers. What this means is that they cannot see what we can see. They cannot see transactions happening across devices. We think they are probably in the order of 30% to 40% more sales they are not seeing because they have not got this cross-device view. By giving them the user graph as-a-service, we can feed into their own tools, this ID, so they can see what is going on, and ultimately, by building that trust and transparency, they can give us credit for those sales. The third and final benefit, in some ways, is even more profound. If I think about advertisers, as Eric said, they are now competing with giants. In the most part, that is Amazon. It is the companies that have strong internal networks, massive scale effects inside their own businesses. For our clients, what they need is a third party platform, a technology platform, who can provide them with the benefits of a broader network. That unleveraged scale affects across everything that we are doing across all our client base. For us, that means building products and services that share and expose that network to our clients. For user graph as-a-service, in many ways, the first step for us on that journey is the first example of a product we are going build that allows our clients the benefit and the strength of our network and deepen, over time, those relationships.
Obviously I would prefer insider buying over insider selling. Most of the selling has been pretty consistent through 10b5, although it did accelerate after last quarter’s beat. I’m not sure if this is due to a depressed outlook or the insiders are simply taking some money off of the table in a volatile stock.
|Entry||11/21/2016 12:37 PM|
I agree that if users increasingly go to apps with a specific product purchase in mind CRTO won’t be as useful. This is also true on desktop. To have value, Criteo must be able to drive incremental sales that wouldn’t have been generated without the advertisements. This means CRTO is reliant on the “research” phase when potential customers are comparing products, which gives them the chance to close a sale or drive an accessory or alternative to what the customer was originally researching. I still think most people do this kind of research before completing purchases in app, which gives Criteo a chance to generate incremental sales for clients. At their investor day, Criteo gave the stat that 40% of eccomerce transaction happen across devices. Comscore says 53%-57% of ecommerce sales occur across devices. This suggests to me that people still do the kind of research necessary for Criteo to have value for clients. It’s also important to note that mobile has grown from less than 35% of Criteo's revenue ex-TAC in Q315 to 57% in Q316 and in app advertising revenue ex-TAC grew over 300% last quarter, making it between 5% and 7% of total revenue ex-Tac, so I don't think the company is being left behind by mobile. I think the concern is less about technology and more about consumer behavior. Will consumers start becoming more assasin like and immediatly buy things that come to mind, or will they continue to research potential purchases beforehand? If it's the former we should be worried about every tech company that generates revenue through ad sales.
|Subject||Re: Re: Mobile|
|Entry||11/21/2016 04:57 PM|
Thanks for the reply - good points.