Influence Attribution in Marketing is an Inexact Science…At Best

The old adage “walk a mile in my shoes” is something that no marketing professional must ever forget. It’s a happy coincidence that the word “path” lies inside “empathy.” You shouldn’t flatter yourself that you understand customer needs with any accuracy unless you’re (vicariously, at least) on the same decision journey as them.

Customer decisions depend on the entire path–experiences, circumstances, actions–from the past to the present moment–and on the unpredictable path that awaits them. So I’m a little bit skeptical of any assertion that marketing-lift attribution can be boiled down to an exact science (see here, for example). If we were able to isolate the definitive factors driving any specific person to take any specific action, the social sciences would simply become social engineering and we’d all be levitating blissfully in “The Matrix” before too long. We may know the physical pathways that customers follow to come to your store or the physical clicks they use to search for and purchase your products online, but do we truly know the decision path that they follow in their heads and hearts?

None of this is to diminish the importance of attribution analysis in marketing campaigns. After all, if you don’t try to isolate the broad factors–media channels, advertising messages, page impressions, customer clicks, agent engagement behaviors, social influencers, etc.–that contributed to a purchase decision, you’re just throwing your marketing investment into a black hole of blind faith. Maintaining a “720-degree customer view” is critically important if you want to grasp all the factors–from external behaviors down to internal propensities and desires–that might be driving their decisions.

However, many of those factors are wild cards that vary erratically in the evolving moment in which we all live. For example, we can scarcely predict what one or more “influencers” in our overcrowded social-network circles are going to say and do next, but that may be the decisive factor on what you and I do next.

Nothing’s truly scientific if it’s not testable and replicable under controlled conditions.

You know that some complex blend of factors–often called the “customer journey”–is what drives behaviors, but it’s a fearsomely difficult string of events to isolate in any scientific way. Your data scientists do their best to track the virtual “breadcrumbs” of data that trace each customer’s peculiar path of events, experiences, and interactions. They can build sophisticated attribution models with as much journey-relevant data as they can lay their hands on. And they can build journey-contextual models that fit historical patterns to a T.

But unless they can test alternate models in real-world experiments inside your marketing campaigns, your data scientists can’t truly confirm the predictive power of those models. Conducting real-world experiments is as much a trial-and-error art as a test-and-confirm science. Your marketing specialists should work hand-in-hand with attribution modelers to vary campaign factors across channels, demographics, time, geographies, and other key parameters. To the extent that they can frame those factors inside plausible journey scenarios, all the better for the predictive power of your attribution models. But in doing so you begin to realize that attribution is more validly applied to the journey as a whole, rather than to any specific click, ad, or other event within that journey.

However, no matter how comprehensively you map the customer journey, the social-influence wildcard will keep rearing its ugly head. You can never hope to define a clear predictive path through a constantly-evolving influence field in which new people enter constantly, old ones become discredited, and all cultural assumptions, values, memes, manias, and practices continue to evolve. Due to the extremely dynamic nature of social influence, it’s not truly predictable, replicable, or testable under controlled conditions.

Sure, you can in theory bribe (aka “incentivize”) the leading influencers to say nice things about your brand. And you can in theory do A/B testing on marketing lift from the “incentivized influencer” scenario vs. “non-incentivized.” But the fluid nature of social influence in today’s dynamic economy–with new influencers rapidly eclipsing established influencers–means that the “controlled conditions” requirement (i.e., static pecking-order of influencers) for true scientific testing will rarely be satisfied. In addition, the chief influencers may be largely “offline”–family, friends, neighbor, colleagues, etc.–and are themselves responding to a shifting blend of “offline” vs. social-media influencers.

Influence attribution is more of an art than a science, and it’s only partially amenable to data science. Where journey modeling is concerned, it might be more valid to frame influence in the online economy as a random walk through a evolving influence landscape than as a discrete journey with fixed mileposts.

Of course, you might think you can use social graph analysis to map the customer decision journey, per this article. But graph analysis is primarily useful only at the social level of influence, interactions, and relationships, not at the individual level of how each of us is experiencing it all. In an influence-attribution context, social graphs are about as useful as Google Earth when what you really want is to know precisely what a particular motorist is seeing through their windshield at this very moment.

You can influence your customers best by being one of them and one with them. You should model their journey as a path of decisions that intersects with your path of empathetic engagement. The journey should be framed as a customer path that is experiential in its desired outcomes, contingent in its nitty-gritty details, and situational in its decision process.

Face it. Influence over the customer buying decision might never be something you can engineer or control. It might never be measurable in any precise way through A/B testing. And it might never reduce to an exact science. And that’s probably all for the best.

Customers want to trust the influencers in their lives. If you’re trying to “engineer” influence, you’ve corrupted its integrity and customers will soon realize you’re trying to manipulate them.

The bottom line is that influence-engineering run amok is the polar opposite of empathy. Customers don’t want you walking their walk if all you’re doing is trying to put your hands in their pockets.

Tags: , , ,

No comments yet.

Leave a Reply