Below are my notes on Criteo (CRTO), a company in the wild world of online advertising. I haven’t finished my detailed financial analysis yet but the company is growing fast and is profitable, although selling for a premium markup. It was a Stock Advisor recommendation on Tom’s side couple months or so back. I have a smallish position in it …
Ideally, you want to read this along with Part II, where I have tried to present a high-level view of the ad tech world.
Understanding Criteo - Part II: Notes from Annual Filing
NASDAQ: CRTO (ADS each representing one ordinary share)
Sector: Information Technology - Advertising agencies
Criteo has started selling its solution in France in 2007 and soon expanded into Western Europe. They started selling their services in North America in 2009, and expanded into the Asia-Pacific market in 2010. In August, 2012, Criteo entered into a strategic relationship with Yahoo! Japan.
Criteo is in the business of delivering personalised advertising, which I believe is a term that is broader than “re-marketing” or “re-targetting”. With “re-targetting” advertisers are targeting users (i.e., visitors to websites) with ads of services/products they have recently seen or perused but not purchased with the hope of translating these window shopping instances into purchases. With personalised advertisements, the idea it would appear would be to go one step further by trying to understanding the users (i.e., understand the users tastes & preferences, their recent browsing activities, and their purchasing habits) by mapping these patterns into models based on data from a very large sample of users using machine learning and data mining algorithms, with the express goal of facilitating ad placements that users are most likely to engage with.
The following bit from the annual filing makes this clear:
For example, in July 2013, 27% of the products sold to consumers to whom we displayed our advertisements were products that such consumers saw on our advertisements and had not otherwise viewed on the applicable client’s website in the 30 days prior to purchase.
So, how exactly does Criteo make all this magical ad placement happen?
- When a new client signs up with them, code is installed on the client’s website that allows Criteo to track visitors to that website.
- Using data collected from the client website and other data sources (more on these later), Criteo’s predictive algorithms to model the likelihood that a user will engage with a client’s advertisement.
- Criteo accesses available ad spots (“inventory” of impressions) through its direct relationship with 8000 publishers, access to real-time bidding ad exchanges, and other ad networks
- The Criteo Engine determines the price to bid for available impressions and if the impression bid is won displays a customised advert on the ad spot.
- The time since an ad spot being available to it being won and served an appropriate ad is under 150ms.
- The results of the campaign, such as whether the user clicked on an advert or made a purchase, are fed back into the Criteo Engine each time an advert is displayed; this is of tremendous value as this data set with “ground truth” (i.e., data where we know what happened when an ad was served based on a set of parameters) can be used to learn what works and what doesn’t, allowing Criteo to improve the accuracy of its production and recommendation engines.
Based on the above description, it would appear that Criteo is essentially a Demand Side Platform (DSP) as it is primarily oriented towards buying ad spots and serving adverts for its clients. The following from Criteo’s most recent 20F Annual Filing nicely summarise this aspect:
We offer our clients an integrated technology platform that enables comprehensive campaign management and execution and includes a unified and easy-to-use dashboard and a suite of software and services that automates key campaign processes. As a result, we reduce unnecessary complexity and cost associated with manual processes and multiple vendors, delivering efficiencies even as campaigns grow in size
Criteo isn’t just a DSP though. It does have extensive publisher side relations, which has the look & feel of a Supply Side Platform (SSP). Criteo has access to a vast swathe of ad inventory through its direct relationship with over 8,000 publisher partners, and they also have a leading presence on real-time-bidding display advertising exchanges. These 8,000 (mostly small- to medium-sized) publishers use Criteo’s inventory management system called Publisher Marketplace (PuMP). The direct partnerships with publishers is useful as it often provides preferred access wherein Criteo is able to select and buy inventory on an impression by impression basis in real time that a publisher might otherwise only sell subject to minimum volume commitments. Further, in some cases the inventory is available to Criteo before it is available to other SSPs or ad exchanges. The strategic relationship with Yahoo! Japan, which gives Criteo privileged access to its advertising inventory for delivering personalized display advertisements is an excellent example of its success.IMO, this aspect of Criteo’s business is pretty important. A strong relationship with publishers indicates that publishers are happy with Criteo’s ability to monetise their inventory, which in turn indicates that there must be plenty of ad buyers waiting to lap up this inventory. It speaks to Criteo’s ability to serve more meaningful or should I say impactful adverts.
The proof is always in the pudding, right, and Criteo’s approach does seem to provide the desired results. For the year ended Dec 31st, 2013, Criteo delivered targeted advertisements that led to approximately 1.9 billion clicks, which resulted in $9.7 billion in post-click sales. Criteo defines a post-click sale as a purchase made by a user from one of their client’s websites during the 30 day period following a click by that user on an advertisement they delivered for that client. Criteo’s clients are certainly happy with the results as reflected in the 90+% client retention rates.
Key Competitive Advantages
Machine-learning based understanding of consumer intent & behaviour
The following is from Criteo’s annual filing:
We have access to two types of differentiating high quality data: (1) valuable consumer purchase behavior data, including products that a consumer has recently looked at or purchased; and (2) our own operating data and insights, which we have accumulated through our experience in delivering over 500 billion internet display advertisements. Substantially all our clients grant us access to detailed consumer purchase behavior data through integration with their websites. We only use the data from each of our clients for the benefit of that specific client’s advertising campaigns and do not sell or otherwise share this data with other clients or third parties. The power of our solution compels our clients to share this valuable data with us, which they would otherwise not typically share. Our own operating data includes insights from user responses to each individual advertisement that we serve, which we use to continually improve our performance. The scale and breadth of our data is constantly growing as users interact with our clients and as we deliver more advertising impressions. For example, in 2013, we analyzed approximately 5.7 trillion ad impressions and delivered approximately 590 billion advertisements.
Instead of rephrasing the above, I will say that the algorithmic advantages translate into cost advantages. The accuracy of their prediction/recommendation algorithms allows them to charge only when users click on an advert, which according to the annual filing is not the common denominator for the display advertising world. In the display advertising world, unlike the search advertising world, the traditional pricing model is to charge if an ad is displayed. So in this sense at least, Criteo is disrupting the market.
The relationships built with 6,600 clients and 8,000 publishers is allowing Criteo to operate a highly liquid marketplace. As long as the number of clients and publishers keep steadily growing, we should be happy, as it means Criteo is making very good headway into gaining market share. This can be sort of have a network effect. More publishers, bring more clients, more clients bring more publishers and so on.
Data-driven virtuous cycle.
This one I will quote verbatim from the annual filing as I don’t really have much to add:
Our solution provides significant benefits to advertisers, publishers and users, which we believe creates a virtuous circle. As we attract more advertising budgets and publisher inventory and deliver advertisements, our data assets grow, enabling us to deliver even more precisely targeted and personalized advertisements and generate more sales for our clients. As our ability to generate sales improves, we believe more businesses will use our solution and increase their advertising spend with us. This, in turn, will enable us to increase advertising revenue for our publishers, further building our publisher network and enhancing our access to their advertising inventory. We expect this virtuous circle will continue to fuel our growth.
Proactive role on privacy advocacy. This I think is quite important and I didn’t see it listed as one of their strengths but I believe these guys have made some good moves to address privacy concerns. The following snippet from the annual filing is worth a reading:
In 2009, we became one of the first companies to broadly include a link in the advertisements we deliver, which gives access to clear, detailed and user-friendly information describing why a user is seeing an advertisement, as well as prominently describing our service and data management practices. In addition, we provide users with an easy-to-use and easy-to-access mechanism to opt out of receiving advertisements
we deliver or being tracked by us either for all campaigns or for a specific client or time period. We believe that this user-centric approach in addressing privacy matters empowers users to make informed decisions on the use of their data. We also actively encourage our clients to provide greater transparency and information about the collection and use of data.
Criteo’s client base consists primarily of companies in the online retail, classifieds and travel segments. These companies range from large, diversified e-commerce companies to many smaller regional companies. Here’s a sample of listing of their clientele: 3 Suisses, Airbnb, BonPrix, CDiscount, Expedia, Gmarket, Gumtree.com, Hankyu Kotsusha, Hokende, Hotels.com, L’Oréal Paris, La Redoute, Lenovo, Lotte, Macy’s, MakeMyTrip, NetShoes, Nissen, Orange, Rakuten, Recruit, Sarenza, Staples, Tiger Direct, Travelocity, and Zalando. My understanding is that the client base is sufficiently diversified and that no single client accounts for more than 5-6% of their total revenue.
In 2011, 2012 and 2013, our largest client represented
5.4%, 5.2% and 5.1% of our revenue, respectively, and in 2013 our largest ten clients represented 18.4% of our revenue in the aggregate.
What are the Growth Opportunities?
Increase market share in existing markets and get into new markets such as Asia-Pacific region & Eastern Europe.
Expand verticals served beyond the three main verticals of retail, travel and classifieds. As such, their solution is generic enough to be applicable in other verticals such as automotive,
telecommunications, consumer goods and finance.
Criteo lists the following as competitors: Amazon.com, Inc., eBay Inc., Google Inc., Conversant, Inc. and Yahoo! Inc. as well as smaller, privately held