Contagious digests the most interesting and relevant research from the world of advertising and beyond, because there’s just too much to read and too little time

Finding creative drivers of advertising effectiveness with modern data analysis /

By John Williams, Nicole Hartnett, Giang Trinh. Published in the International Journal of Market Research

Give it to me in one sentence.

If you want your TV ad to boost short-term sales, your best bet is to introduce the product towards the beginning and the brand name towards the end.

Give me a little more detail.

The researchers (two of whom hail from the Ehrenberg Bass Institute for Marketing Science) were interested to know which elements of video ads were most effective at driving sales.

They applied the so-called ‘modern data analysis paradigm’ to a set of 312 TV commercials, provided by a single FMCG company, that spanned 60 different brands across five markets. The data also included information about how the ads affected sales, measured by monitoring households and recording whether they purchased a product after being shown (or rather, had the opportunity to see) an ad for it.

The researchers checked the ads against a list of 158 different variables – marking whether humour or music was present, how long it took for the product to appear, etc – resulting in almost 50,000 observations. They then applied 317 different statistical models to the data, to discover which of the variables consistently correlated with sales increases.

The different models frequently disagreed about which variables had the largest effect on sales, which proved the researchers’ point about the limits of relying on a single model to interpret data. But the best machine learning and artificial intelligence models consistently found that the timing of the appearance of product and brand name mattered a lot.

Introducing a product (or packaging) early in an ad was linked to improved sales, and so was introducing the brand name (or logo) later, in the middle or final third of the spot. Ads that showed the brand name or logo for shorter periods of time (relative to the duration of the whole ad) also performed better, which probably ties into the finding about later introductions.

‘Based on our findings, we propose an empirically derived theory that showing the product early serves as a contextually important cue to set up introducing the brand,’ write the researchers.

Why is this interesting?

This study is all about the method. The researchers write that previous studies mostly relied on one analytical model to interpret data, which leaves doubts about whether the findings reveal the true drivers of short-term sales, or whether they tell you more about the assumptions and specifications of the model itself. The modern data analysis paradigm employed by the researchers tries to solve this quandary by unleashing loads of analytical models on the same data – and in this case it threw up some unexpected results. The previous studies on creative effectiveness tended to pinpoint things like humour and music as the biggest drivers of sales, not the length of time it takes for the product or logo to appear.

The researchers also say that their study shows that machine learning and artificial intelligence methods ‘generally give better fit and prediction than the classical method used previously to analyse these data.’ But they are still wary of being too prescriptive about it all.

‘We do not wish to imply that these creative variables guarantee sales effective advertisements,’ they write. ‘There is no foolproof formula for how to make a great, sales driving advertisement. The findings do, however, highlight a handful of creative tactics (eg, early product introduction, later branding) that proved more consistently effective than other tactics when subjected to varied analytic testing.’

The researchers were also careful to state that creative elements like humour and mascots still play a vital role in drawing people’s attention to TV ads.

Any weaknesses?

The researchers point to their relatively small sample size and the fact that they employed people (rather than machines) to code the ads. But from our perspective, the biggest weakness is that we’re taking a lot on faith when it comes to the analytical models, which we are in no position to critique.

Where can I find the whole report?

Here and it’s free.

We don’t just write about best-in-class campaigns, interviews and trends. Our Members also receive access to briefings, online training, webinars, live events and much more.