Transparency Where It Matters

gillespie-criteo-ddt“Data Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is by Jaysen Gillespie, Director of Analytics and Business Intelligence at Criteo. It is a response to an earlier AdExchanger column by Tal Keinan, CEO of AdExtent (Read it: “DSPs Vs. Personalized Retargeters. One Step Forward, Two Steps Back?“)

“Transparency” is the hot buzzword these days. However, most vendors are applying the word to the wrong concept. A recent post from Tal Keinan, CEO of AdExtent, extolled the virtues of extending the reach of DSP-based advertising to include personalized retargeting in order to provide transparency. While I agree that personalized retargeting is a critically important component of any integrated marketing plan, I would argue that comingling a DSP-based upper funnel “prospecting” campaign with lower-funnel retargeting is a step backward in terms of real transparency.  There are a number of reasons why starting with high-quality personalized retargeting is in fact the “two steps forward” the industry needs.

Transparent ROI

What has always mattered in performance marketing is return on investment. If you are getting a true 8:1 ROI with Vendor X and a 5:1 ROI with Vendor Y, it’s irrelevant how Vendors X and Y obtain internal margins with their respective technologies. Paid search vendors run high gross margins but still provide a solid ROI to clients.

Transparent Vendor Margins are the Real Sucker’s Bet

Transparent vendor margins, on the other hand, provide a client with a false sense of security about performance. In the internet of the past, where inventory was bought and sold on a CPM (cost per impression) basis, comparing vendors on a cost-plus-markup basis may have provided some insight into vendor efficiency. In today’s world of real-time user valuation and advanced predictive modeling technologies, cost-plus pricing allows vendors to execute profitably, even when their solutions are sub-par. If vendors can continue to garner a 20% margin based on inferior solutions, outdated technology will prevail over true performance.

Holding vendors to the highest possible ROI standard not only makes more sense for the marketer, but also allows vendors to do what they do best: leverage advanced technology, publisher relationships, and the entrepreneurial ecosystem that undergirds the advertising technology space to drive performance.

Transparent Methods Prove ROI

Personalized retargeting vendors know they have to prove their worth: ROI can’t be obscured by blended results with other marketing initiatives. The most structurally sound method for proving ROI is to split the cookie pool of site visitors into two mutually exclusive groups, with one group getting the personalized retargeting and the other group getting no personalized retargeting.  The split must be truly random so that the composition of each pool is the same in every respect, and no other retargeting outside of the vendor under study should be applied to either cookie pool.

Once the cookie pool is split, it’s easy for a vendor or client to track performance (conversions or sales revenue) in each group. The lift in revenue per unique user from exposed to unexposed groups provides the basis for calculating incremental revenue. True ROI is then a simple matter of total incremental revenue from the exposed group divided by cost. Strong vendors can perform this split and insert client-visible tracking codes, such as DART trackers, to bring full transparency to the process.

Many Users, Different Habits

A DSP can play an important role in bringing potential new customers to an advertiser’s website. However, once the new prospect has arrived, specialized retargeting companies are the most efficient solution for “closing the deal” with that new prospect. All-in-one solution providers run the risk of becoming the ERP systems of the online ad world (general solutions to specific problems). The online advertising ecosystem acts as a team, allowing the client to select the solution that meets their needs most efficiently for each type of customer the marketing strategy chooses to address.

One ROI Hurdle Rate For Each Unique User

Leading edge marketers are realizing that the days of having a marketing “budget” are long gone.  Instead of allocating $X to retargeting and $Y to prospecting – or relying on a third party to shift dollars from one group to another, performance-driven marketers are stipulating what ROI or Cost-of-Sales needs to be hit and then go uncapped as long as results are exceeding those goals. Financially, this makes much more sense: If your business model is to allocate 10% of revenue to cost-of-marketing, you would want to spend as much as possible on a solution that generated a 12:1 ratio of revenue to cost. In effect, today’s budget-constrained marketers are doing just this by repeatedly extending the budget for solutions that scale and are above the desired return rate. This isn’t restricted to COS metrics, as many marketers have other metrics they measure success by. As long as they have a clear feedback loop and at least a basic attribution framework, this same approach can be extended in most cases.

The value that DSPs and other prospecting tools bring to advertisers is most clearly registered when the costs and benefits of upper funnel initiatives are separated from lower funnel results. Keeping a “best of breed” approach to marketing spend creates this level of transparency around the performance of each vendor. Such a separation also allows third-party attribution solutions to correctly attribute conversions to their respective vendors across the entire sales funnel.

Follow Jaysen Gillespie (@jaysengillespie) and AdExchanger (@adexchanger) on Twitter.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. Alejandro Correa

    I take issue with many points in your article but two in particular:

    – ” it’s irrelevant how Vendors X and Y obtain internal margins with their respective technologies.”

    Distilling any service soley to “performance” opens the door to all kinds of fraud. I’m sure that mortgage backed securities were great investments until the truth came out. Not saying anyone here is commiting fraud (Criteo is a very reputable company in my experience), but if that’s what we play at, we’re inviting disreputable players who know how to game the system to come in, fool the clients with shiny metrics for long enough to cash out and skip town.

    – ” …value … is most clearly registered when the costs and benefits of upper funnel initiatives are separated from lower funnel results.”

    I disagree wholeheartedly. Value would be most clearly registered if adspend could be viewed holistically, and the contributions of upper-funnel tactics could be tabulated when determining attribution for a sale. Looking at CPA for a retargeting line separately from a tv spend, or search, or other display, is like saying that the cost of an apple is equal to the cost of picking it from the tree. Before that apple could be picked, it had to be planted, watered and cared for…this is essentially what upper funnel media does. Retargeters and other lower funnel channels simply harvest sales…an important function, but hardly representative of all the work that goes into actually making revenue.

  2. Hi Alejandro,

    Thanks for your thoughts on the article.

    I think that we are actually in what I’d call “violent agreement” on the core concepts: performance has to be presented fairly and transparently in order for it to work as a yardstick. For that reason, I’ve outlined in the article some thoughts around how to create that fair and transparent type of analysis, such as using controlled AB testing where clients and vendors have full visibility into cookie pool selection and results. Retargeting presents a unique situation where you can split the pool of users directly – because, by definition, such users have already been to the site. Once the users are split without bias (and this can be done by a neutral third party like a tagging solution vendor), it’s straightforward to conduct the AB testing needed to isolate the impact of the retargeting.

    On your second point, again, we agree on the core concept: marketers need transparency around the costs and performance of each part of the marketing funnel. Retargeters do indeed benefit from TV, radio, and upper-funnel digital, as all those marketing efforts serve to build the potential audience for retargeting. The controlled AB test again addresses this need: by creating a control group that gets all the other upper funnel marketing, but not the retargeting, and comparing the performance of the retargeting-exposed group to the control group, a client can isolate the uplift that is due solely to the retargeting effort. Once that specific uplift is known, it can be used as an additional data point to help clarify the impact of other marketing vehicles.

    • Alejandro Correa

      Thank you for thorough response, it makes sense and I think I understand what you were trying to get at better now. The part of your article that I was reacting to concerned it being “irrelevant” how performance is achieved as long as it was achieved. Putting that aside, I’m still not sure I’d be 100% comfortable with an approach that measured retargeting individually from other other tactics:

      assume the client only cares about CPA at the line item level and has a last touch attribution model. It seems like it will always be in the retargeter’s best interests to serve low cost impressions at a high frequency to try to harvest attribution for conversions. In the best case scenario, the campaign may not be as efficient as it otherwise could have been because of the unneccessarily high frequency. In the worst case, the overall eCPA of the campaign becomes less efficient as the eCPA for retargetting becomes more efficient. If the client doesn’t understand this (because the tactics are irrelevant) then they will mis-judge the value of their media as a whole.

  3. As long as we are circling around the topic of transparency, I might as well through my $.02 in as well…

    In reference to your title, “Transparency Where it Matters” – my thesis is that it actually very much matters when the actual client is asking for it. Many existing partners of yours that we talk to disagree wholeheartedly with the statement that: “… it’s irrelevant how Vendors X and Y obtain internal margins with their respective technologies.”

    While performance and ROI are absolutely essential, being able to achieve that performance combined with transparency appears to be where the market is headed as we evolve from partnering with blackbox oriented solutions optimizing their own margin in addition to their client’s goals.

    So, again in the spirit of transparency, maybe a good place to start would be disclosing your margins as this appears to be an irrelevant issue in your mind.

    Jason Kelly
    CEO Sociomantic

    • @Jason: great comment. Echoing this, What if all publishers and display RTB exchanges were to require that vendors disclose media costs versus fees? The value that vendors are creating via “optimization” pales in comparison to the value of the retargeting data (owned by the advertiser). Time for advertisers to start asking more questions.

      • Retargeting data certainly has value but the whole point of the A/B test is to quantify the value of the Ad serving (which includes all of the optimization). A properly designed control would have a random sample of the retargeting population. The conversion rate of the control group quantifies the value of the retargeting data. The difference between the conversion rate of the test group (i.e., the retargeting uses that saw an Ad) and that of the control group is the value of the Ad serving and optimization. If A/B tests are run on all campaigns and the results are passed through to the client, that is amazing transparency. The logic behind exposing margin is like asking your shoe retailer to expose their costs before you decide the shoe’s value to you (and thus the price you’re willing to pay). Most economic transactions are made by exchanging equal amounts of value between the actors. A properly executed A/B test should clearly show the client exactly the value they are receiving. If the value is greater than the costs, the rational person would take the deal.

        Now without an A/B test, the client is likely buying back a lot of the value they already own.

  4. Roger Hemmingworth

    “The logic behind exposing margin is like asking your shoe retailer to expose their costs before you decide the shoe’s value to you (and thus the price you’re willing to pay). ”

    You’re right. That is why Nike often hides the fact that its costs are so low – and avoids talking about working abuse and child labour in 3rd world countries.

    A shoe is something tangible, something you can hold in your hand. For the average person, it is easier to gauge the value or a shoe than an impression. What is an impression worth? What is the true value? Is it the end result, what it achieves? Perhaps – but why not be transparent about the original cost? There is no reason to hide this other than to hide the fact that certain organisations have disgustingly high margins and are just looking out for their own figures rather than their clients’.

  5. In a DR focused campaign, which applies to most, if not all, retargeting, an impression is worth:

    (the naive version)

    Probability of Conversion After Ad Exposure * Value of a Conversion.

    (the non-naive, A/B tested version)

    (Probability of Conversion w/ Ad Exposure – Probability of Conversion w/o Ad Exposure) * Value of a Conversion.

    Everyone on the client/agency side should have this equation memorized, and they should hold every DR campaign they run to this yardstick. And if they run retargeting, they should definitely A/B test.

    Now I qualified my argument to DR focused campaigns. I agree that in a Branding campaign cost-plus might be the more appropriate model.