LONDON/NEW YORK: Marketing mix models are looking increasingly slow and out of date as agile attribution models become more widely used, but neither type has all the answers, according to two leading industry figures who advocate integrating the two.

Writing in the current issue of Admap, Alice Sylvester and Jim Spaeth of the Sequent Partners consultancy outline How to combine the best of marketing mix and attribution models to optimise effectiveness.

There has been "a swift and very striking change in marketers' expectations in ROI media performance measurement" over the past three years, they note, with mix models now being seen as "too slow, too macro and too backward thinking for the boots on the ground".

The granularity and speed of attribution modelling, on the other hand, is very appealing.

But the two are quite distinct, with different strengths and limitations; results rarely agree and marketers don't know which results to follow, making it imperative that a way is found to combine the two.

"There may be an elegant, holistic solution down the road, but the best practice we have seen to date is layering attribution beneath marketing mix," say Sylvester and Spaeth, as they advocate a technique that takes the best aspects of each and links them to business outcomes and measures.

This process involves four layers with brand KPIs, the metrics that guide the business, at the top.

Under that marketing mix modelling reveals the drivers of the brand's marketplace results, while a third layer of granular mix models can "tease apart the effects of different executions in a campaign".

The fourth, bottom layer of the stack is attribution modelling, which is still a work in progress, but which the authors expect will develop with self-adjusting algorithms and machine-learning artificial intelligence to provide better accuracy and more sensitive forecasts.

The next challenge in attribution, say Sylvester and Spaeth, is to break down the silos between digital and traditional media: since attribution operates solely in the digital media silo, the models are unaware of the effect of traditional media, the rest of the marketing mix and the brand itself.

"This isolation results in gross overstatements of the impact of digital media," they point out.

But when the problems with attribution modelling are overcome – and there's a long way to go – they argue that marketers will shift from measuring ROI to managing ROI.

Data sourced from Admap