Understanding Criteo - Part I

To understand the complex and sometimes murky terminology of the world of online advertising, we need to start with some very simple basics. And, before we delve into anything, let me caution you that this overview is in neither comprehensive nor authoritative. This is simply an overview that gives some colour to the Criteo (CRTO) investment thesis.

Publisher, Advertisers, and Impressions

Let’s start with some basic terminology:

  • Publisher is the owner of online services where ads can be placed.
  • Ad spot or impression is the spot for inserting advertisements on the online service.
  • Publishers are the ‘sellers’ of ad spots.
  • Advertisers are the ‘buyers’ of ad spots.

Online ads can be broadly classified into two categories: search and display ads. Search ads are ads that are displayed on the top or side of a search result; this is more or less a monopoly in the Western world with Google’s big dominance over search. Display ads are ads that are displayed on Websites These can be banner ads with some text/images and can also be video ads. CRTO is playing in the display ad market.

Ad Exchanges

Not too long ago, the buying and selling of ads was relatively straightforward. Buyers of ads interacted directly with the sellers of the ad spots. However, with the exponential growth in the Web and Internet usage, this direct transaction model became very quickly inefficient. Publishers were often left with unsold inventory of ad spots. If we think about the “long tail” of websites, there’s just no way by which sites with moderate traffic would be able to move their inventory efficiently. This situation gave rise to “advertising exchanges” (or simply ad exchanges) for buying and selling ad inventory.

An ad exchange is similar to a stock exchange in that it’s a marketplace for advertisers and publishers to respectively buy and sell ad space, often via real-time bidding. Effectively, an ad exchange is a large pool of ad spots (or impressions) which have been made available by the publishers and now advertisers can buy ad impression across a range of publishers. Ad exchanges provide liquidity for buying and selling of ad spots. Buying of these ad spots can be achieved very quickly, often under 150ms. Often, this speed of execution is critical as some users quickly move across websites and advertisers feel that they need to grab the users attention before they navigate elsewhere. Below is a list of the some of the top ad exchanges today:

  • DoubleClick Ad Exchange (Google; this is the biggest one)
  • Appnexus (probably second behind Google; handled about 16 billion ad buys per day in 2013)
  • Yahoo! Right Media Exchange (RMX)
  • OpenX ad exchange (privately held)

Supply Side Platforms (SSPs), Demand Side Platforms (DSPs), and Data Management Platforms (DMPs)

When combing through 10K’s of businesses in the world of online advertising, you will find the word “platform” attached to almost anything one can conjure up. I ‘m not kidding you, and IMHO the distinction between some of these platforms (and also ad exchanges) seems blurry. Nonetheless, let’s try to make some sense of this madness.

SSPs are tools that enable publishers to sell their ad inventory for the best possible price. SSPs connect the publishers to multiple ad exchanges, ad networks, and DSPs, essentially exposing their inventory to a large swathe of potential customers with the goal of trying to maximise the earnings one can get from the ad inventory. Sounds a lot like a big ad exchange because as a concept it is one! SSPs are offered by Google (AdMeld by DoubleClick), AppNexus, Yahoo (Right Media), and Pubmatic, OpenX, and many others.

DSPs are the advertisers counterpart for SSPs. DSPs are systems that serve ads on behalf of the advertiser in realtime based on specified rules. Those rules are based on the data provided by advertisers and by inventory sources (Exchanges and SSP’s) when an impression is offered for auction or sale. DSPs may use various data sources to perform behavioural targeting, for example use tracking data to determine if the audience segment and/or content should be targetted. To summarise, DSPs often offer a centralised system for managing ad campaigns, deciding which ad impression to buy based on a host of data, some of which may be proprietary data to the DSP platform. Examples of DSPs include Digilant, Invite (acquired by Google) MediaMath, Rocketfuel, and Criteo (although we can argue that Criteo is a mashup of DSP + DMP + ad network).

DMPs can be thought of as a data warehouse. These house data from first party (i.e., clients/advertisers) and third-party sources (e.g., things like data from loyalty programs, surveyors etc). Data can be from online as well as offline sources. With advertisers buying ad spots across multiple exchanges and SSPs, and using sometimes multiple DSPs, DMPs play the role of making sense of all the various data feeds with the goal of optimising ad buying decisions. Examples of DMPs include Adchemy, Datalogix, Demdex (now owned by Adobe), Digilant, Experian, and Mediamath.

Targeting Techniques

Broadly, there are two types of advertisement campaigns, “audience response” and “brand lifting”. In the former, there are specific goals that are desired (e.g., increase sales, increase website traffic etc), while the latter is harder to measure but is essentially about building/enhancing a brand reputation. Irrespective of the the of campaign, there’s always a specific target audience for the campaign.

What are the common techniques used to reach the target audience?

Contextual targeting - This is ad placement based on the content/context of the publisher of the ad (e.g., online brokerage ad on the business pages of online news)

Demographic targeting - This one is self explanatory.

Keyword targeting - This is similar to contextual targeting except that the ads are relevant to certain keywords, either on the pages where ads are displayed or based on keywords entered by the user for a search.

Lookalike targeting - The idea here is to classify users into groups based on how users reacted to ads. Then, when a new user is to be served an ad, determine which group this user is mostly likely to belong to and serve ads that were most effective for the chosen group. This is a useful tool for finding new customers.

Re-targeting or re-marketing - The idea here is to try and convert visits that didn’t result in a sale into a sale. For instance, suppose you search for a vacation package on Expedia, but don’t proceed to buy the package. Expedia could then have ads served to you, opportunistically with the goal of bringing you back and converting that missed sale into a sale. Criteo specialises in re-targeting/re-marketing.

Privacy and Cookies

The following paragraphs from Rocket Fuel’s most recent 10-K nicely describes the privacy challenges being faced by the online advertising industry. Privacy is the achilles heel of this sector. Since Rocket Fuel gave a vey good overview of the challenges, legislative and otherwise, I will simply post that part verbatim:

Internet users’ online activity generates a vast amount of data, such as advertising viewed and responded to, and advertisers’ websites visited, and is valuable to advertisers seeking to reach an optimal audience. Online it is possible to serve advertisements to potential consumers based upon inferred interests. These interests may be inferred in part based on web-browsing history. The use of web browsing history to inform advertising purchase decisions is commonly referred to as “interest-based” or “online behavioral” advertising. Advertisers are willing to make a greater investment in, and pay a higher rate for, digital advertising when this interest-based data can be used to inform decisions about purchasing advertising impressions to reach desired consumers.

The use of interest-based advertising on the Internet has come under scrutiny by consumer advocacy organizations and regulatory agencies in the U.S. and abroad that focus on online privacy. More specifically, these groups have voiced concern about the use of cookies and other online tools to record an Internet user’s browsing history, and the use of that information to deliver advertisements online based on inferred interests of the Internet user. Because we rely upon access to large volumes of data, including web browsing history, primarily through cookies, it is essential that we monitor developments in this area in the U.S. and globally, and that we engage in responsible privacy practices.

The “Do-Not-Track Online Act of 2013” was introduced in the United States Senate in February 2013. If a “Do Not Track” web browser setting is adopted by many Internet users, and the standard either imposed by state or federal legislation, or agreed upon by standard setting groups, requires us to recognize a “Do Not Track” signal and prohibits us from using non-personal data as we currently do, then that could hinder growth of advertising and content production on the web generally, and limit the quality and amount of data we are able to store and use, which would cause us to change our business practices and adversely affect our business.

That’s it for now. The next part will cover the Criteo investment case in some detail.

References and Further Reading

1)The Evolution of Ad Tech, MIT Tech Review, Sept 2013. Online: http://www.technologyreview.com/view/518551/the-evolution-of…

  1. Driving Marketing Results with Big Data, MIT Tech Review, Nov 2014. Online: http://www.technologyreview.com/view/532551/driving-marketin…

  2. WTF is Demand Side Platform? Online: http://digiday.com/platforms/wtf-demand-side-platform/

  3. WTF is a data management platform? Online: http://digiday.com/platforms/what-is-a-dmp-data-management-p…

  4. WTF is Supply Side Platform? Online: http://digiday.com/platforms/wtf-supply-side-platform/

  5. Online Advertising, Wikipedia. Online: http://en.wikipedia.org/wiki/Online_advertising

  6. Use remarketing to reach past website visitors and app users, Google. Online: https://support.google.com/adwords/answer/2453998?hl=en-AU

  7. Rocket Fuel 10K Feb 2014. Online: http://investor.rocketfuel.com/secfiling.cfm?filingID=104746…