Criteo has provided a point by point rebuttal to some — but not all — of the criticisms levelled at it in a recent Seeking Alpha story. However, it has shared that information with investors rather than customers.
The email (see below) provides some insight into the kind of heat Criteo is taking from its investors as a result of the report. (It’s share price has dipped a little in the week since the publication of the Seeking Alpha article, however it has recovered in recent days) Publicly, the company has merely issued a blanket denial, most recently in a short statement on its investor relations page.
In an email from Criteo to Morgan Stanley and obtained by Which-50, the company runs through a point by point explanation of its response to some criticisms.
However none of the material addresses the specific concerns Which-50 raised in recent correspondence with the ad tech company, which were also published in the Seeking Alpha piece.
Those concerns were garnered from court documents relating to the now-resolved legal dispute between Criteo and rival SteelHouse.
Given the audience — investment advisors and by extension their clients — the material focuses more on business model issues than on specific technical details.
For instance, in response to the suggestion in the Seeking Alpha article that its Return on Ad Spend (ROAS) metric “… is opaque itself, with Criteo providing little colour to investors on how it arrives at ROAS figures,” the company says:
“The calculation of ROAS is 100 per cent transparent. Clients calculate ROAS for all marketing vendors by dividing sales attributed to that vendor by the spend for that vendor. For example, if Criteo drives $3,000 in last-click sales in a client’s web analytics platform and the client spends $300 on Criteo marketing, then the ROAS is 10:1.”
In the email, Criteo argues that the calculation is not only transparent but conducted by the independent standard web analytics tools themselves. “Furthermore, in a typical ‘competitive bake-off,’ clients will run Criteo and another vendor during the same period of time. The client, not Criteo, performs their own ROAS calculation to determine whether Criteo or the other vendor delivered superior performance. Criteo has won approximately 90 per cent of these comparisons.”
Through an agent, Criteo has told Which-50 that they will not be responding further on the subject. We will, however, happily keep asking questions.
For instance, it would be good to get explanations for the following from the SteelHouse litigation (SteelHouse says it analysed Criteo’s performance across a number of sites and discovered these figures – although bear in mind it was a hostile litigant at the time):
- Over 50 per cent of Criteo’s clicks (as measured by SteelHouse) had no original source (6x greater than industry average), suggesting potential “bot” traffic;
- 3.6 per cent of Criteo’s users generate 25 per cent of its clicks;
- 44.9 per cent of Criteo clicks come from “clusters” — 13x the industry standard;
- SteelHouse found one IP address that clicked on a Criteo ad 20,647 [times] across 96 of all 99 advertisers queried.
One final point: there are quite plausible scenarios where the above figures are true, and yet are not the result of deliberate actions by Criteo (and assuming no wilful blindness on its part). Under such a scenario, Criteo is still open to criticism for not being sufficiently robust in monitoring traffic on its network.
Which-50’s view is that Criteo (and all ad techs frankly) should commit to transparency around such issues.
In the meantime, here is the email Criteo sent to Morgan Stanley. We believe the author to be from the Investor Relations group at Criteo. Given the nature of this story we feel it is only fair to republish it here in full.
Most of you have received requests from investors about the SeekingAlpha blog post that was published yesterday and already reached out to us. Please find Criteo’s response below. In addition, we question the credibility of such source given that they are short sellers.
Please let us know if you have further questions on the topic.
The commentary published on Seeking Alpha by an anonymous contributor includes a number of unsubstantiated and false claims directed at Criteo and our business model. Criteo’s 90%+ customer retention rate (maintained for over 20 quarters) is ample proof that there is no validity to the claims in this article. We do not use adware. We make money only when our clients do, because our business model is aligned to driving their sales. We rigorously monitor suspicious click activity automatically to ensure that only genuine clicks are charged to our clients. Criteo’s systems are regularly audited by external auditors.
Article: Criteo would only do so [buy inventory] if its algorithm believes that the traffic on Seeking Alpha’s web site is likely to translate into clicks because Criteo’s “revenue” is ultimately dictated by click-through rate, not by impressions (this is important).
Response: Criteo’s revenue is dictated by click-through rate and conversion rate. Clients set the cost-per-click (CPC) based on hitting a target cost of sale, which requires that clicks from Criteo convert into e-commerce site sales. The article’s implication that Criteo benefits from sending low-quality or fraudulent clicks shows a misunderstanding of Criteo’s business model. If Criteo sends clicks to a client’s e-tail site that don’t convert, the client will respond by lowering its CPC, thereby reducing Criteo revenue. Ultimately, Criteo and client interests are 100 per cent aligned because Criteo drives post-click e-commerce transactions, not clicks.
Article: In other words, Criteo’s algorithm is designed to only keep ads on a web site if the click through rate is strong.
Response: The above statement is factually false. Criteo’s algorithm is designed to purchase any display where the expected cost of sale of the associated e-commerce transaction is below the client’s target cost of sales. Criteo might thus choose to buy ads on a web site where the click-through rate is relatively weak, but where such clicks have a higher-than-average conversion rate once they land on the e-commerce web site. The actual bidding formula used by the Criteo engine contains, among other terms, the product of the predicted click-through rate and the predicted conversion rate, ensuring that Criteo only buys traffic that converts into e-commerce sales.
Article: Criteo touts its “return on ad spend” aka ROAS as the holy grail of all metrics, used in competitive bake off situations to win and retain clients. The actual calculation of this ROAS is opaque itself, with Criteo providing little colour to investors on how it arrives at ROAS figures.
Response: The calculation of ROAS is 100 per cent transparent. Clients calculate ROAS for all marketing vendors by dividing sales attributed to that vendor by the spend for that vendor. For example, if Criteo drives $3,000 in last-click sales in a client’s web analytics platform and the client spends $300 on Criteo marketing, then the ROAS is 10:1. This industry-standard calculation is not only transparent but conducted by the independent standard web analytics tools themselves. Furthermore, in a typical “competitive bake-off,” clients will run Criteo and another vendor during the same period of time. The client, not Criteo, performs their own ROAS calculation to determine whether Criteo or the other vendor delivered superior performance. Criteo has won approximately 90 per cent of these comparisons.
Article: And if you still trust Criteo to get attribution calculations “right” — then perhaps this recent news release from Facebook will make you rethink your belief:
Response: Attribution is performed by independent web analytics systems, such as Google Analytics, Adobe (Omniture) SiteCatalyst, and IBM’s CoreMetrics, that the client chooses. E-commerce sales are attributed based on attribution rules applied to all traffic, such as “last-click attribution.”
Article: #1 — Surprise History — How Can A Business With Presumably Low Visibility Always Crush Revenues?
Response: Because Criteo sells post-click performance, as measured transparently by client chosen web analytics systems, Criteo experiences a 90 per cent client retention rate which results in strong revenue predictability.
Article: Based on the litigation, it is apparent that Criteo’s main beef with SteelHouse was around utm codes relating to desktop last click attribution.
Response: utm_ codes are used for sales attribution across desktop, tablet, and smartphone devices. The vast majority of transactions from mobile devices are delivered via the mobile web, not via In-App purchase. As a result, the author mischaracterises Criteo’s concern as “desktop only.” While Criteo cannot comment on the litigation settled with SteelHouse, Criteo invites all readers to consult those publicly available filings to understand that this statement is inaccurate and incomplete.
Article: From a practical perspective, this [referencing the fact that conversion rates are lower for mobile devices] means that in order to generate the same amount of clicks on a mobile device, you need to buy far more impressions (people are simply clicking ads far less frequently on mobile versus desktop).
Response: The above is factually inaccurate. Mobile devices do experience lower conversion rates compared to desktop. However, such a statement does not imply that the click-through rates are also lower for mobile devices. The author has no basis to conclude that “you need to buy far more impressions” to generate the same number of clicks on a mobile device.
Article: We also believe that mobile CPMs are actually higher than desktop CPMs, removing the possibility that higher volume buys could be offset by lower price points.
One would therefore naturally assume that in the near-term, Criteo’s margins should have come down as Criteo has been forced to go out and buy more impressions — increasing the “CPM” side of its two-sided margin equation, in order to keep the CPC side intact.
Response: For the week ending Nov 30, 2016, Criteo’s CPM on Desktop was $3.21. Criteo’s CPM on Mobile was $3.14. There is no basis for the author to conclude that Criteo needs to pay a higher CPM on mobile devices or that our margins would be affected by the overall market dynamics. Criteo evaluates each user, on each device, in real time, which is a vastly different approach from that of many marketing vendors. It is thus impossible to apply overall market trends to Criteo’s specific inventory buys.
Article: If CRTO really is a cutting edge technology company, then why is its R&D spend so low when compared to its comp set?
Response: The author is not comparing apples to apples. Google and Facebook have no direct costs for most of their internet traffic. Thus, it would be a more relevant financial ratio to consider the percentage of Revenue ex-TAC that Criteo spends on R&D. If we take 40.4 per cent of the 1,323M in “Gross Revenues” to yield 534M in Revenue ex-TAC, the R&D spend of 87M represents 16.3 we cent of total Revenue ex-TAC. This 16 per cent figure aligns with that of Google’s spend.