How the right data can help you to break out of the digital cult

by Nigel Hollis

Some of my colleagues expressed concern over last week’s post which likened online advertising to a cult. But what else do you call the unthinking faith that many marketers have in the power of digital advertising? Sadly, data has a lot to do with reinforcing that faith, but the right data can also help marketers break out of the cult.

In the original article by Jesse Frederik and Maurits Martijn referenced in my post they describe an occasion on which eBay stopped buying search ads on the MSN Network for the keyword “eBay” Instead of traffic dropping off, people simply switched from using the paid link at the top of the page to the ordinary one in the regular search terms. eBay was paying $20 million a year for a search term that was bringing no additional traffic. A subsequent three-month test with Google simple produced a similar result.

eBay is not alone in discovering that a lot of search advertising is buying them very little in the way of additional traffic. Witness the case of Adidas, which, as the result of a breakdown in Google AdWords, discovered that traffic and revenue did not fall when paid search stopped. However, the problem of assumed effectiveness goes well beyond paid search.

In this video, Adidas global media director, Simon Peel, contrasts the results of attribution models and market mix modeling to highlight that misattribution does not just apply to paid search but also referral sites, online display and email marketing. Adidas’ econometric modeling found that video advertising – online and off – and social ads were more important drivers of sales than was otherwise believed.

The fundamental problem is that all too often digital advertising targets people who were intent on buying the brand anyway (the selection effect). That intention may have arisen from brand exposure that happened incidentally at some point in the past before people even knew they want or needed to buy the brand’s category. Or it may have arisen because the person already owns or uses the brand and needs to replace it. Convenience metrics like clicks, views and downloads do not discriminate between people who already want to buy your brand and those that do not.

None of this is to say that all digital advertising is wasted. It is not. But relying on convenience metrics to measure success is not going to ensure effectiveness. Instead marketers need to do what eBay and Adidas did, test and learn. From the big picture perspective, understanding how all marketing investments create predisposition to buy in future and trigger sales now, econometric modeling provides a good road map provided the model includes brand metrics that measure predisposition as well as sales data.

At a more granular level, controlled tests to identify advertising effects (rather than selection effects) are easy to implement by conducting a test and control comparison to see whether a campaign is creating that all important predisposition to buy. And, at the most granular level, if you want to test whether an ad will work well in different contexts there is a solution for that as well.

Digital advertising does not have to be a blind act of faith but, of course, the first step to breaking with the cult is to admit you might have a problem.