Attribution Modelling: Beyond the Last Click
Attribution modelling has been in the news recently, with brands such as Boden crediting this approach to measuring RoI with transforming their view of how to invest their digital budget.But what precisely is 'attribution modelling'? And what role does it have to play in the marketing planning process for premium and luxuy brand marketers?Put simply, it’s about moving away from the approach which attributes all of the credit for a sale to the activity which generated the 'last click' and towards apportioning that credit to all online activity that played a part in moving the consumer down the purchase funnel. It's about taking a more holistic view of digital marketing spend, and giving more credit to those activities - like display advertising and generic search terms - which are unlikely to be driving the final purchase, but are clearly playing an important role in the purchase process nonetheless.But it's more than that. At its most sophisiticated it's also about understanding what types of activity generate what types of customer. For example, shouldn't activity that drives new customers get more credit than activity that retains existing ones. And shouldn't activity that drives higher margin business get more credit than activity that drives lower margin business?The premis is simple and logical. The problem is, attribution modelling is not so simple to implement.To begin with, research needs to be undertaken to understand whether attribution modelling is relevant and what sort of model should be applied. For example, if a brand isn't employing a mix of online channels to drive demand, it's unlikely that attribution modelling can add much to its existing 'last click' analysis. The same applies if the purchase process is simple - attribution modelling is not going to add much value for all the effort involved. However, this is unlikely to be the case for most premium and luxury brands, where the nature of the purchase and the sophistication of the audience are likely to mean a long and complex buying process is going to be the norm.However, once the research is complete, there are plenty of tricky decisions to come.The first of those is to do with timing, namely how long is the product's buying cycle? Activity that happens before this 'timing window' needs to be discounted from the analysis. For example, should searches or display banner views that took place 90 days prior to purchase be given credit or are these events unrelated to the final purchase decision because they were too distant from it?Secondly, decisions have to be made as to how to attribute the credit. Should the first interaction with the brand's marketing activity be given more credit as it prompted the customer to start the purchase journey? Or should the last, that prompted the purchase? Or should all stages along the journey be treated equally? How much extra credit should be given to activity that was successful in generating new customers or more profitable customers? Should someone clicking a natural search result be given more weighting than a display ad view?The third challenge is implementation - unpicking the carefully constructed systems that have supported last click models of RoI and rebuilding them to support attribution modelling. And the complexity deepens the more systems that have to re-configured.So is attribution modelling worth all that effort?Well, it’s certainly not perfect. The whole process involves its fair share of educated guesswork and mistakes are going to be made. And marketers can't fully understand what credit every stage in the process deserves because they can't understand every customer - 2 identical purchase processes could have 2 very different triggers for purchase depending on what stimulated those individuals. And of course, we're only attributing a fraction of the spend - offline activity such as press or magazine advertising isn't part of this process as yet - although there's no doubt that attribution models that incorporate offline channels are being looked at.But some understanding is better than none. We all know the process of purchasing premium and luxury goods is much more complex than one click = one purchase. So it’s time for our methods of measurement to catch up with our knowledge of the purchase process.Attribution modelling may be the start of a long road, but it's better than the dead end that last click analysis represents.