In my last blog, I laid out some facts that call into question the extensive effort many organizations put into attributing individual customer sales to individual marketing touch points via common attribution methods. To summarize, Suresh Pillai, head of Customer Analytics & Insights for Europe at eBay, showed that all reasonable attribution algorithms led to effectively the same aggregate credit to each marketing lever and also the same credit as a random method.
If in many cases it isn’t possible to achieve a result different from a random allocation, is all lost? The answer is no. While the detailed customer data utilized for attribution may not be as useful as expected for traditional attribution purposes, it still has a lot of value. Let’s discuss how.
LEVERAGING ATTRIBUTION DATA
Data and, therefore, information, is valuable. I assume that anyone reading this will agree with that assertion. At the same time, any given piece of information may not be relevant or helpful for any specific purpose. In other words, information has value when placed in the right context. Suresh’s conclusion (and my own) as outlined in my April blog simply states that customer level data about which touch points preceded a purchase do not necessarily help with the process of allocating credit for a specific sale for a given customer back to those touch points. However, there are other uses for the information.
Another key point of Suresh’s talk, which I did not address last month, focused on how the patterns of touch points that preceded a purchase can provide useful insights in other ways. In other words, it is useful to know the fact that a customer first saw an email, then later performed a search, and then made a purchase. It ends up that there are certain mixes of touch points that outperform others when it comes to driving sales.
Suresh is not alone in this assertion. I also saw one of my colleagues from Teradata, Yasmeen Ahmad, discuss the exact same concept at an event that I also spoke at in London earlier this year. She discussed how segmenting customers based on the mix of touch points each customer utilized led to some very interesting insights.
As luck would have it, I also saw my friend Justin Cutroni, Google’s Analytics Evangelist, make the same point when we both spoke at an event at a Wharton Business School event a few weeks back. When I see three experts share the same suggestion within weeks of each other based on completely different work streams, it gets my attention.
THE PATH FORWARD
The important takeaway is that detailed data about how a customer works toward a purchase is, in fact, very useful. Rather than using it in the traditional way for attribution, use it to segment customers based on the combination of touch points that lead to purchase (or lack thereof). Once customers are segmented according to these interaction patterns, substantial insights can be achieved about how customers behave. It is also possible to apply very precise costs to each customer’s interactions since you’ll know what you had to pay for each touch point that the customer utilized.
Really, this distinction comes down to a common issue with analytics. Namely, any time that you aggregate or summarize data some information is lost. By going straight to attribution, we effectively aggregate the data for each customer from the outset. However, by maintaining the precise pattern of interactions over time for each customer, we retain additional information that is lost once the data has been aggregated.
My conclusion last month was a bit of a “downer.” It pointed out the gap between perceived and real value of common attribution processes, and it also didn’t provide any ideas about how to redirect efforts to achieve a positive outcome. This month, I want to wrap up with a positive conclusion and a high impact way to move forward.
By all means, continue to collect the detailed information that feeds traditional attribution processes. By using this data in a different manner than is typical, you’ll be able to derive real value. Identify each customer’s unique path to purchase (or lack of purchase) and then group customers on that basis. You’ll likely learn a number of things both about your customers and your marketing efforts. In addition, you’ll be able to assign costs and match that to revenues in order to assess what mixes work best.
When I see experts from eBay, Google, and Teradata all reach the same conclusion independently, it makes me take notice and reconsider my own thinking. I suggest you similarly reconsider your thinking and commit to exploring how the approach outlined in this blog can help your organization.