Multi-touch campaign attribution is an imperfect solution to a vexing problem. Though perhaps better than simpler attribution methods, multi-touch attribution by virtue of its over-the-shoulder rear view will never match, with 100 percent accuracy, an action to a reaction.
Some form of multi-touch attribution is needed to understand increasingly complex, multi-touch, multi-channel customer journeys. But marketers are far from measuring and improving multi-touch campaigns in an optimal way.
Dynamic customer journeys lay bare the shortcomings of first-touch, last-touch, evenly weighted distribution or other traditional attribution methods, which create incentives for doing multi-touch attribution. First, traditional methods use haphazard associations. If a customer makes a purchase on a website directed from a Google search, for example, a last-touch attribution model may not acknowledge or credit the influence of the ad the customer had engaged with on YouTube earlier that day, or that two days prior the customer had browsed the same product on a branded website.
Likewise, the seemingly randomized associations may skew the cause-and-effect decisions further downstream, from channels and campaigns to offers, messages, content and any number of factors. When marketing dollars are tied to these factors, which they eventually will be, it is easy to see how marketers end up increasing chaos or contention in their allocations vs. approaching performance marketing.
In fixing multi-touch attribution, marketers have the opportunity for a more accurate representation of customer journeys and customer response to interactions, one that measures the effectiveness of a campaign or channel with a higher degree of certainty.
Post Hoc Fallacy
Before examining how a customer data platform (CDP) can help with multi-touch attribution, we should recognize some difficulties with multi-touch attribution that are inherent in the “rear-view mirror” approach of trying to understand the causes of a customer action that’s already happened.
First, multi-touch attribution is subject to a logical fallacy: Since Y follows X, Y must therefore have been caused by X (in Latin: post hoc ergo propter hoc). If a consumer purchases an outfit after seeing an ad, it may be easy to assume that seeing the ad triggered the purchase. But perhaps the buyer’s daughter, thinking her mother would like the outfit, put it in an online shopping cart and shared the link. The “recorded and measured” sequence of events does not in and of itself show the reasons for the purchase.
Another limitation of multi-touch attribution is that an allocation across touchpoints may fail to account for multiple touches within a specified allocation. A brand that attributes 70 cents of each purchase dollar to an advertising campaign and 30 cents to a direct mail piece may decide to boost spending on advertisement. But if a campaign consists of multiple advertisements, perhaps one is effective, and the others are instead driving customers away.
Having a clear understanding that multi-touch attribution is not a perfect solution may enable marketers to better appreciate the broader picture, viewing multi-touch attribution as a vehicle for (and a product of) experimentation and optimization rather than simply for resource allocation.
Dive into Experimentation, Optimization
A website product recommendation engine offers a good example for how multi-touch attribution should be used in an experimental context. A rudimentary engine may limit offers to best-sellers, regardless of a customer’s behaviors or preferences. Machine learning models that create special offers for thousands of audiences may result in higher sales, but a data-driven brand will still likely do an enormous amount of a/b testing in order to assess a new recommendation engine’s effectiveness.
Similarly, with the advertising example, using multi-touch attribution as a vehicle for testing and optimization may provide a good view of the root causes that contributed to a customer’s action. Rather than a straight-up resource allocation (70/30 advertising vs. direct mail), this broader use of multi-touch may help explain why customer sentiment moves in a particular direction.
A CDP & Multi-Touch Attribution Software To The Rescue
Using multi-touch attribution in an experimental/optimization framework is where a CDP comes into play, as long as that CDP offers an experimental framework. Understanding at a basic level if one channel works better than another is a start, but vetting a calculation to understand why it works better, to consider and adopt or discard external forces and variables – that magnifies the power of multi-touch attribution.
The Redpoint rg1 customer experience platform provides such an experimental framework to enable marketers to easily set and test an endless number of combinations (messages, offers, channels, business rules), foregoing assumptions by accurately pinning down the entire customer journey.
rg1 provides three core capabilities to optimize multi-touch attribution. First, it includes tools for accurately matching data from anonymous interactions to data from known interactions, and accurately matching offline and non-digital activities with online, digital activities for a more robust picture of every customer interaction throughout an unknown to known journey. Providing complete, accurate, relevant and timely information is an important first step in doing attribution calculations.
Second, multi-touch attribution is only as good as a CDPs ability to present cleansed, accurate, normalized, time-stamped and relevant data to the analytics framework – within the CDP or as a separate capability – that does the multi-touch calculations. This is critical to allow for broad experimentation, such as running multiple attribution models simultaneously to truly understand customer behaviors.
The third important capability of rg1 is that it collects and shares attribution results. Making attribution calculations, after all, is only half the battle. The other half is making the information available and putting it into the hands of the people who are going to do something with it. If, for example, a business is tracking KPIs based on attribution in a management system, it’s necessary to make the underlying data and the attribution calculations available to the management system at the cadence the system requires to operate. Likewise, when the CMO sees attribution figures that show the success of particular channels, the CMO should have confidence that the dashboard is accurate, up-to-date, and provides the sources and calculations for the attribution.
In summary, to engage in multi-touch attribution a customer data platform cdp should be able to bring all the data together, make it available to the appropriate analytical engine and calculations being used, and make both the underlying data and the attribution calculations available to the system of record for tracking attribution.
With Change Comes Opportunity
Multi-touch attribution is a timely subject because of the phasing out of third-party cookies by Firefox, Safari and – next year – Google. These have been a significant source of attribution for many years for measuring advertising effectiveness, and both marketers and Ad Tech and Martech vendors are scrambling to find replacements or change practices.
The demise of the tracking cookie presents opportunity for marketers to take a fresh look at multi touch attribution tools. Rather than just reallocate attribution in a haphazard manner, an examination of the underlying goals for multi-touch attribution may convince an organization of the advantages of adopting an experimental/optimization approach. The right CDP can help.