Attribution: How To Break The Holding Pattern

adamberke“Data-Driven Thinking” is a column written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Adam Berke, president at AdRoll.

Whether you’re a brand marketer, an agency person or at an ad-tech provider, you’ve probably had several awkward conversations about attribution.

Pundits, myself included, have predicted this to be the “Year of Attribution” – for the past few years. Yet, when it comes to actual decision-making on performance marketing strategy, last click still carries the day.

The industry has made some strides in the right direction. For example, it’s now widely understood that last click is an inaccurate and arbitrary measurement methodology that leads to poor decisions.

There is also no shortage of technology to help marketers move beyond last-click measurement. No system is perfect, but options that provide a more complete picture are certainly readily available. Even Google, which has been criticized for benefiting from and perpetuating last-click measurement, is adapting by providing better multitouch attribution tools.

So if knowledge and technology aren’t the problem, why does last click still have such a strong hold when it comes to measuring performance marketing campaigns? Consider two phenomena – and a solution.

1. Disproportionate Focus On Pricing Models

When executing performance marketing campaigns, there is a simple equation that must be solved. You need: media spend, a margin that reflects the value created by a technology provider and the resulting return on investment (ROI), which needs to exceed the advertiser’s goals.

Media costs + vendor costs < target acquisition cost (ROI)

Unfortunately, a disproportionate amount of time is spent focused on the left side of the equation, rather than trying to make the right side – the ROI – higher.

Pricing model and negotiating the vendor margin can influence the left side. The right side can be influenced by more strategic and durable factors, such as campaign strategy, optimization plan, creative and core technology like bidding algorithms.

Focusing on the right side, or ROI, frees up attribution to take a more prominent role in the execution, since it decouples pricing model from measuring results. When I ask marketers why they care about pricing model, they often say that performance-based pricing, such as CPA or CPC, aligns the vendor with their goals.

This can be a dangerous misconception. All these models do is encourage the vendor to optimize around an attribution model upon which they charge and control. Pricing model doesn’t change the original equation.

Will the vendor continue running a campaign if they don’t make a margin? Of course not – they’ll call back to renegotiate pricing if the arbitrage doesn’t work out. Will the marketer use vendor reporting as the source of truth? Not a chance – they’ll use reporting from their ad server or, hopefully, a neutral third-party analytics solution.

Who cares what pricing model the vendor employs if the marketer isn’t going to use the vendor’s reporting as the source of truth? Regardless of the pricing model, there is a cost that must be justified based on the marketer’s target ROI, which is measured in an entirely different system. Focusing on the pricing model takes away from focusing on achieving and measuring actual results.

Please note: I’m not suggesting that ad-tech vendors be held unaccountable in contractual terms. However, if an advertiser wants to negotiate a term that aligns the vendor with their goals, instead of pricing model or margin, I recommend pushing on lowering the minimum spend amount. If a customer can leave at any time, that’s the sure way to align the vendor to deliver results on the customer’s terms, based on the attribution model of their choosing.

For example, sometimes a marketer will want to change campaign variables that actually lower performance, such as adjusting frequency caps, running a seasonal flight or excluding publishers. If the vendor is only rewarded for clicks or conversions, they don’t have an incentive to be flexible on those points. Now if the pricing model is based on actual media spending, and the advertiser can pull out any time, will the vendor accommodate those and other requests? Absolutely.

Focusing on vendor margin, just like focusing on margin when buying any product, isn’t a good way to assure quality. It’s quite the opposite. Diageo makes a larger margin on Ketel One than on Smirnoff, but that doesn’t mean you’d rather have Smirnoff in your martini.

Look at the recent succession of ad-tech IPOs. Are the 40%-plus margins demanded by companies like Rocket Fuel and Criteo unreasonable? Not if those margins are indicative of the value those companies bring their customers.

2.  The Attribution Catch-22

As the saying goes, “The road to hell is paved with good intentions.”

The same can often be said about attribution model adoption, resulting in an unfortunate Catch-22. This occurs when a marketing team understands the benefits of more sophisticated attribution but fails to use it to make actual decisions.

This puts media buyers and ad-tech vendors in a tricky situation. As mentioned above, different attribution models drive different campaign executions and optimizations.

I’ll use an example from the retargeting world. In a conversation with a marketer or agency about measurement, all agree that the best way to measure the actual impact of retargeting is to compare the behavior of a group of people who are exposed to the retargeting campaign to a randomly selected control group that sees public service announcements. Measuring the campaign this way ensures that the retargeting vendor goes after conversions that move the needle for the business, instead of just chasing clicks from people who would have converted anyway. Focusing on incremental conversions in this way is almost certainly going to result in higher CPCs and CPAs, but since there’s less cannibalization of free return conversions, the overall ROI is much better.

Everything is great until the end of the test, month or quarter, when someone at the brand or agency goes back to their old last-click model and notices that the CPC or CPA went up.

That puts the media buyer and ad-tech company in an impossible position. If they deliver to the more robust attribution model, they sub-optimize for the more simplistic last-click model. In that respect, knowledge and technology aren’t what’s holding back attribution. It’s purely organizational commitment.

If organizations actually use attribution to make decisions, and let go of pricing models to instead focus on what drives true results, we can finally make this the “Year of Attribution.”

Follow Adam Berke (@adamberke), AdRoll (@adroll) and 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. Great article and you’re totally right about how companies are behind the times still using last touch. I suggest advertisers start having conversations with the attribution platform teams in the space such as Adometry, Convertro & VIQ. Hold out tests on re-targeting campaigns are better than the norm but you’re still only looking at touches within the display channel.

  2. Glad you pointed out the incrementality aspect in your Catch-22 paragraph. The more sophisticated attribution systems do a good job at measuring correlation (i.e. path to conversion), but they do not measure causation. As you allude to in your PSA example with retargeting, one way to measure causation is with carefully designed experiments using test and control groups.

    It is dangerous when organizations implement an attribution solution and then establish a rule, like: “view-through gets 20% credit for a conversion”. At some point in the past, for some particular campaign, a 20% credit for conversions on a view-through may have been appropriate – but assigning fixed rules for fractional credit attribution models is taking a good technology that can help point out correlation and making the leap to causation. This may be convenient, but it is certainly not accurate.

    One way we have been tackling the attribution challenge with our clients is by measuring “all” conversions that are influenced by the campaign – both online and offline – with a closed loop reporting integration. We also focus on measuring the net, incremental impact of campaigns as you described above. One of the weaknesses of online attribution systems is correlating offline sales with online activity and/or cross-device activity. So online attributions systems are challenged at measuring some kinds of conversion activity. Using a closed loop approach to capturing data and a net incrementality experimental design has helped us get beyond debates of how much to credit a particular conversion event (20% or 30% for view through?) and focus on what matters most – incremental new customers.

  3. Adam,
    I agree with your last sentence. But everything else tells me that you and/or your client had some really bad experiences when working with some attribution vendors. Most of the issues that you mentioned above will not happen in a typical attribution engagement, particularly when the advertiser and attribution vendor are well aligned on the marketing goals right from the kick-off, and the attribution vendor consistently tracks those goals over time by comparing attribution-based measurements and traditional last-click measurements side-by-side. Marketing attribution does not deliver ROI in and of itself, but attribution-based insights and the operationalization of its optimization-recommendations typically result in a 10-25% gain in “incremental” ROI on media spend. It’s this incremental ROI that makes the right side of your equation go up much higher for those marketers who successfully adopt attribution-based-optimization. However, the importance of operationalization attribution cannot be overlooked. The most successful marketers are able to automatically deploy attribution-based recommendations to their media buying execution platforms (DSPs, RTBs, DMPs and Exchanges) using automated feeds provided by their attribution vendor. Finally, attribution science must take incremental gains into account from an ROI perspective. The retargeting example you provided should not happen. Advanced attribution science will take care of new acquisitions through retargeting correctly.

    If you would like to know more about how marketing attribution can be done correctly, please contact Visual IQ.