Today’s post is courtesy of Jim Spaeth and Alice K. Sylvester of Sequent Partners.
“ROI.” “ROI.” “ROI.” whined Jan Brady.
Ok, maybe not, but you can’t go too far these days without encountering the term Return on Investment. ROI performance is on everyone’s mind – auto dealers, big retailers, movie companies … all of radio’s advertisers.
But just as we wrapped our heads around the advanced math of econometric marketing mix models, the measurement world changed. Today, the venerable, powerful marketing mix models are considered too slow, too macro and too backwards-thinking for most marketers. They need tools that are more granular and more comprehensive – tools like they have in digital, where they can attribute sales to digital touchpoints and map a consumer’s journey from search to website to reviewers’ blogs to Facebook to Amazon.
Attribution modeling is leaving the digital ecosystem and will play a key role in cross-platform ROI analysis. It’s a very hot topic full of promise and right now, some bluster.
Multi-Touch Attribution, MTA.
With the right data inputs, scientists will be able to assign value to all consumer touchpoints – promotions, price reductions, traditional and non-traditional media – the full marketing spectrum.
There are a lot of challenges being addressed as the industry makes this shift away from marketing mix models and into attribution modeling. Marketing Mix models looks at the effectiveness of marketing elements at the store/week/DMA level – in a more macro, aggregated way. Attribution modeling occurs at the household or device-level – and looks at events and exposures in a much more granular level.
Attribution modeling itself – the assumptions and the math – is changing. Originally, attribution or credit for sales generation was assigned entirely arbitrarily – a priori. What drove the sale – the last click, when someone actually bought something? First click, when they began looking into buying something? All of this work took place without any validation and rationale.
But that’s changing. The best attribution models are based on sound statistical modeling with parameters fitted to real data. The techniques used today range from time-series models to game theory approaches, but providers are not always as transparent as they should be. And the audience data that feeds the models is not always there at the levels the moderators need. We expect this to settle down under increased industry scrutiny in the near term. Ultimately, iterative, self-adjusting algorithms and machine learning – artificial intelligence – and highly disaggregated data will produce the accurate sales forecasts we need.
Streaming radio is well-suited for attribution modeling because exposure data is relatively easy to get. But terrestrial radio is not. Clearly, getting the right data to feed this new measurement animal should be a top industry priority.
But in the meantime, what should you listen for when people talk about attribution modeling?
- Are they really only talking digital? Are they talking about TV and digital? Or are they talking about the whole mix?
- When they talk about radio being in the attribution models, and in the “data stack,” are they talking only streaming radio? How are they thinking about terrestrial radio?
- And when they talk about campaign effectiveness, are they talking about real creative executions/messages or unidentified impressions? What are they measuring?
Those three questions should help you understand where the person on the other side of the desk is coming from. Don’t let radio remain on the sidelines. These new analytics should do a better job of detecting and representing the impact of radio, but radio needs to get in the game.
You can hear more about this topic at the industry’s Attribution Accelerator Forum in New York on November 30th. For more information and a full agenda please visit here. And check back here for more — we’ll have follow-up on the topic after this event.