A New Perspective On Attribution

sephzdarko“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 written by Seph Zdarko, head of attribution initiatives and partner strategy at Quantcast.

Attribution is one of the most critical issues in digital advertising today. Proper attribution provides insights, incentives and controls over ad spend.

However, our current attribution models are flawed; they are too simple and easily gamed, as they measure from only the conversion event and skew metrics and incentives to bias the lower-funnel tactics. Measurement from more than just the conversion event is needed, and adding a second signal of measure further up the funnel is the right next step. If the industry chooses to adopt it, attribution will evolve to a new level of control and effectiveness.

Several of the leading attribution companies are exploring this option, but it requires greater storage and infrastructure to implement, so it needs everyone’s support. It is time to move from one signal of measure to two. This is a simple step that will have profound impact on how we view upper- and lower-funnel programmatic planning. A second point of measure will enable new metrics, insights and incentives for more efficient spend resulting in more conversions and less gaming of attribution.

The skill set and data required for effective upper-funnel prospecting is fundamentally different from lower-funnel retargeting. Ideally the two tactics work hand in hand to maximize total conversions across the full funnel. However, when only measuring from the lower-funnel conversion, they compete for credit and retargeting always wins because it is closer to the point of measure. By creating separate metrics for each phase independently, they now can work holistically together in a more efficient partnership to increase total conversions.

Adding A Second Signal For Dual Credit

Here’s how to do it: Split the funnel in two by using the first site visit as the second point to measure from. The first site visit is a natural delineation point between prospecting and retargeting. Consumers might visit a site multiple times before they convert, but retargeting can’t start until that first visit happens. This provides a clean break in the data and a natural hand-off between tactics. The site visit is the natural transition from upper and lower funnel and it touches every converter. Now we have new perspective and new metrics for prospecting and retargeting independent of each other.


Notice in the above illustration that a conversion is still required in both the upper-funnel (prospecting) and lower-funnel (retargeting) phases. That means credit is only attributed after the conversion. Just getting a site visit isn’t good enough. We still incentivize the converting customer, but now we have two separate incentive structures. Prospecting is about getting someone who has not been to a site to visit and convert; retargeting is getting someone to convert after they visit, but now they work hand in hand without competition and with full transparency.

It is important to remember that the job of your attribution model isn’t about rewarding or punishing vendors or departments on a plan. Attribution is an optimization tool for maximizing conversions. The byproduct of the optimization process is that sometimes budgets are cut, increased or decreased, but the goal should always be towards maximizing the desired outcome.

This measurement approach can produce double-digit increases in total conversions. The reality is that last touch attribution skews the influence of lower-funnel activity for traditional metrics like CPA, ROAS or ad efficiency that often result in overbudgeting and oversaturation with retargeting, meaning fewer conversions.

Let’s look at the example below:


Notice the difference in CPA calculations. In the above scenario the campaign data is the same for both reports; the dual-credit report just accounts for upper-funnel metrics independently from the lower funnel and is able to break out ad effectiveness more cleanly.

It is important to note that the resulting optimizations you might enact would be vastly different based on the conversion and CPA data presented in each report. This campaign example (it has been cleaned for illustrative purposes) produces a 30% increase in conversions simply by moving budget to vendor 1, an upper-funnel prospector, and vendor 4, a retargeter. The phased partnership results in more conversions without any additional budget.

Just budgeting a balance of prospecting and retargeting tactics on a plan isn’t the same as measuring them. Dual measurement is essential for having true perspective into your marketing funnel initiatives, and this two-phased dual-credit approach is the next step.

Marketers can ultimately assign whatever value or use whatever multitouch strategy they want within each phase. It is completely flexible, provides separate incentives and is less easily gamed. While this discussion has focused primarily on programmatic, this model can, and should, be applied across all digital marketing channels.

Adding a second signal of measure provides a natural next step in the evolution of attribution today. Two-signal attribution also lays the foundation for further advancements: adding a third and fourth signal to measure from for potentially even more control and insights. But let’s not get ahead of ourselves. It’s time to give upper-funnel attribution its due respect and split the funnel!

Follow Quantcast (@Quantcast) 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

    I think many agree that attribution could improve but the solution you propose is problematic. Although it would help general adnetworks look more efficient vis-a-vis specialized retargeting, it is unclear to me that the advertiser benefits. By looking at an up-funnel metric, am I really driving incremental sales? A better approach perhaps would be to measure lift, which is not without problems, but may be a better way to demonstrate value.

  2. To the point on – Prospecting is about getting someone who has not been to a site to visit and convert; retargeting is getting someone to convert after they visit, but now they work hand in hand without competition and with full transparency.

    The challenge is with multiple sources driving audience, how will you single out a new visit and attribute to the specific vendor & what will be the incentive for the new visit.

  3. Thanks Alejandro. I agree that measuring lift is an important piece of attribution, but measuring the lift from prospecting independently form retargeting is just as important. the proposal of splitting the the funnel into two phases is is just one piece. For example, I didn’t even mention the additional layer of viewability in this. Many companies have already, or are in the process of, layering in viewability to remove those ads in both upper and lower funnel tactics that were not seen by the consumer. Adding a second signal to measure from is about gaining more insights, not the final verdict. Using regression or game theory as many do is an additional filter to layer in as well. It will be a long time before attribution is solved, if ever, but the goal is to keep improving and let the outcome of the optimizations your attribution model suggests be the deciding factor of what ultimately prevails. This is all about making continual steps in the right direction for the betterment of our industry as a whole.

    • alejandro c.

      I agree with much of what you say but am skeptical of the value of measuring the lift of prospecting indendently of retargeting. Why not do whatever maximizes the total number of conversions for the advertiser? If splitting up the funnel does that, then great. But if there is overall lift at the cost of prospecting “looking” like it is less efficient in a last touch attribution model, then I think its a small price to pay.

  4. Is the first ‘site visit’ a strong signal in all cases? We’ll need to justify that first. This is essentially linking clicks to conversions, but we already know that a high CTR doesn’t always have a positive correlation with a high CVR. The given example looks promising, but one can also imagine opposite cases especially when the automated optimisation comes in.

  5. Not only is the current attribution model flawed, but the premise of this article is flawed as well for any marketer whose products are sold (even partially) through brick and mortar stores and not on their websites. The act of optimizing spend to deliver website traffic (or website KPI’s) is fundamentally misleading if the website isn’t where the sale takes place. Until we as an industry acknowledge this, we’ll continue to not get the full respect of true marketers. We end up focusing on the lowest common denominator because we can do so relatively easily.

    Start with the desired outcome (the sale) and then work backward to identify the various places that transaction takes place (online and offline), then look at the actions (online and offline) that lead to that transaction and begin to measure them. Then consider how each marekting tactic influences these metrics. Then and only then will you have a true attribution model from which to optimize.

  6. I agree that we should not treat apples as oranges and vice-versa.

    An accurate attribution algorithm will consider the difference between retargeting and prospecting and will do it at the most granular level (e.g. placement, creative).

    Moreover, it will account for differences in viewability and factor them into the algorithm so that marketers can understand the true performance of their different marketing activities.

    This is what we already do at Convertro and we are looking forward (together with our friends Seph and his team at Quantcast) to continue educating the industry about the benefits of utilizing accurate, data-driven, algorithmic attribution to get actionable spend recommendations