• Home
  • Uncategorized
  • Marketing attribution analysis: the remarkable (but scary) new world of data

Marketing attribution analysis: the remarkable (but scary) new world of data


|

In a recent “Lunch and Learn” presentation given by an analytics firm we work closely with, I had a revelation that was equally terrifying and inspiring. We are in a remarkable time. New analytical advancements allow us to exploit very deep, complex patterns in our behavior. What’s overwhelming, though, is dealing with the amount of data shoppers generate every day and the incredible array of shopping pathways it can take, involving many marketing touch points and many devices. This raises some tough questions. Which touch point gets credit for a sale? What pathways drive most of the sales? Is one combination of touch points better than another at conversion? Historically, our ability to answer those questions was limited and often required the use of best judgment. But a new generation of analytical methods, called marketing attribution analysis, promises to supply more precise and predictive answers.

What is marketing attribution analysis?

Marketing attribution analysis is the application of statistical method that assigns value to each of the marketing touch points along a shopping pathway, a value that reflects the relative strength of that marketing touch point in driving a sale. The most popular method today is simplistic “single touch” attribution. This method assumes the first or the last touch gets credit for the sale, but we know that’s an inadequate and inaccurate assumption. Simplistic “multi-touch” methods are an improvement but many are still lacking in their rigor. They attempt to assign credit by judgment, by equal weighting, by time (the longer the time between a touch point and a sale, the less credit it gets), or applying a regression-based approach. While better, their effectiveness quickly dissipates when the number of touch points becomes numerous or the data vast. The advanced multi-touch methods showing the greatest promise are those that utilize advanced statistical models (such as Markov chain modeling) as their analytical foundation.

Why do we need it?

The reason we need this more advanced approach is that we no longer have any control over how and when a consumer is exposed to our message. Today’s consumer is in complete control of their shopping experience. They have more ways than ever to engage with a brand. Studies increasingly reinforce that a consumer will use multiple channels on the same shopping journey, depending on the context they are in at a particular moment. With just five different channels by which a consumer can interact with your brand, there are more than 3,000 different pathways a consumer can take before purchasing.

Given that most marketing plans will incorporate more than five marketing channels and executives will continue to demand the ROI on marketing expenditures, we must be able to optimize the use of our marketing resources to achieve the greatest return. We will need to be able to:

  • move resources away from unproductive channels
  • know what channels are contributing to conversion
  • assess the impact of different interactions (or combinations of interactions)
  • and predict the next step that will generate the greatest return

The more sophisticated multi-touch marketing attribution models will enable these decisions.

Coming soon: how it works

In the coming weeks, I’ll be sharing some additional perspectives from our analytical partner about how this new approach works. I’ll forewarn you – this is not for the faint of heart. It takes some intellectual fortitude and resolve, but it’s worth it. The companies (particularly agencies) that take the time and effort to develop a competency in this area will be at a significant competitive advantage. They’ll be smarter about their customers, more efficient in the use of their marketing resources and ultimately more profitable. I think any C-level executive would think that’s a good thing.

Leave a Reply

Your email address will not be published. Required fields are marked *