Note: I’d like to attribute article ethos to several industry sources from the likes of Adexchanger.com to Quantcast
…Attribution modelling is a great idea; particularly for display advertising.
What can sometimes happen though is that display advertising is either gamed, mismanaged / sub optimally optimised or still treated in a noughties fashion (aka last click attribution).
Example 1: If the advertiser is working to a last click model; the majority of the budget could be spent on retargeting consumers who were probably going to convert anyway (achieved as a result of sophisticated DSP technology that predicts close to checkout behaviour).
Example 2: A way of gaming attribution modelling that has a focus on crediting view throughs is to buy lots of useless cheap display inventory (usually small formats, those below the fold or even those hidden in an iframe or interstitial) rather than by using metrics such as 3M (see my 3M’s post for details).
Example 3: What about those visitors who have been to the checkout, noted you accept voucher codes, gone to a website for the code and then been targeted with a RTB banner and thus cookied (which again IMO added little value to that customer journey)
5 steps to improving your display advertising attribution
1/ Split out your targeting into prospecting and retargeting. Understand the role of each and target + KPI appropriately.
2/ Lay down strong KPI’s (i.e. this is certainly not going to be a reach number 😉 that are in turn baked into a robust & relevant multi click attribution model (see my dedicated attribution post on this topic from a couple of years ago).
The model you devise will likely still value view through conversions and windows but does not over indulge these metrics as is sometimes the case e.g. abandon a last click wins 365 day post-view through window model for a multi click attribution model (where the model is really laboured upon for relevancy to your industry) with a post view 14 day window.
3/ Coupled with the last point; use strong measurement targets that further shows up obvious cheeky gaming tactics e.g. a KPI for measurement might be that all display ads have to fully load and at least 80% of the banner be visible in the browser window for more than 1 second.
4/ Test activity in an A/B fashion and see, over the same time period, how parts of the site are £returning without the assistance of display prospecting and/or retargeting. i.e. a PSA (test and control) lift test.
5/ Consider your reward model. Look to reward key players in your display advertising funnel in a different fashion. e.g. rather than rewarding your agency on a % of spend basis what about flat fees with a robust SLA written with detailed activity expectations.