DMexco spotlight: the new challenges of marketing attribution
One of the most interesting talks I attended at this year’s DMexco was on the subject of marketing attribution. AdRoll President & CMO Adam Berke’s presentation ‘Demystifying attribution in a Multi-Device World’ tackled the tough question of how to attribute marketing success now that customer journeys are split across multiple devices.
The problem of marketing attribution
It’s important for marketers to know how much revenue each of their activities are generating. Marketing attribution attempts to match up each value-creating action of the customer (e.g. making a purchase, requesting a quote, signing up to a newsletter) to a particular marketing channel, campaign or piece of content.
This has always been a tricky task, as customers frequently come into contact with several marketing touch points before parting with their cash. For example, if a customer sees a billboard advertisement for a product, then sees the same product advertised on social media, and finally performs a Google search for the product before making a purchase, how do you attribute that revenue? To the billboard, social media or organic search? Or a combination? (How would you even know about the billboard view?)
The multi-device challenge
The advent of mobile and tablet devices has made this problem even more pronounced. While attribution of purely online activities could conceivably be fairly comprehensive when customers were restricted to using a desktop computer with cookies enabled, tracking users across multiple devices, some of which are less than friendly to traditional tracking methods, is much more complex.
Another ramification of this complexity is ‘analysis paralysis’. Marketers are now bombarded by so much data about their customer’s behaviour across devices and sessions that knowing where to start their analysis feels impossible.
Many marketers make use of marketing attribution platforms to ease the burden of analysis and help them put an ROI figure against their various activities. But no platform provider can deliver a genuinely 360-degree view of customer interactions, and so they’ll often be biased towards the statistics and metrics that they can gather most comfortably or which push you towards marketing solutions that they sell. This can also be true when using an external agency to help with attribution: there’ll often be a bias towards the agency’s own campaigns/assets.
The last click fallacy
Lots of marketers currently rely on last click attribution methods. This is where all the revenue from a particular visit is attributed to the last known marketing channel that the customer touched upon. In the billboard-social-Google example from above, last click attribution would result in all the revenue being attributed to organic search.
As you can see, this method is inherently flawed as it often ignores all the other touch points in your customer journey, not to mention people moving between devices. Adam gave a good analogy for this is in his talk: if you were to hand out a 10% off coupon to everyone currently stood in a supermarket queue, and if you were using last-click, you’d (falsely) attribute all the sales made to the coupon-wielding customers to the coupon.
According to Adroll’s report, nearly half of marketers are currently using last-click attribution, with a similar number using the even less reliable first-click model.
Lack of skills and confidence
Skills and training are key barriers also to carrying out more complex attribution analysis. With technology moving so fast it’s hard to find people with the full set of skills and knowledge required to implement good attribution models. Interestingly there are also some regional differences, with the UK predominantly taking a ‘learn on the job’ approach, whereas France and Germany do much more in the way of formal training.
You can read lots more about marketing attribution by downloading the full Adroll report, ‘The State of Marketing Attribution 2016’. It doesn’t present any easy answers to the above challenges but does give an excellent overview and presents advice on choosing the right attribution models and technologies.