There’s a lot of buzz about "attribution modeling" in the online media world right now. There’s also a lot of confusion about what it is, how to do it, and what it’s worth. If we can make one small and succinctly stated contribution to the discussion of "˜how’, it’s this: having access to an online panel makes it a heck of a lot easier to do it well.
For context, "˜what’ we are talking about is giving credit for a target outcome (generally a conversion or transaction) to the media exposures that preceded it. An advertiser who records 20,000 purchases during Q1 would like to know which campaign strategies contributed to those sales, and how much each strategy contributed.That knowledge can help optimize future marketing strategies.
Advertisers have the best view of outcomes — the anchor for any attribution analysis — but on its own an advertiser’s view is often limited to its own customers and site visitors. An attribution model that looks only at site visitors can discover which media strategies correlated with conversion/transaction, but won’t reveal which placements did the most to turn unknown consumers into visitors and customers.
Ad servers help by contributing a view of media exposure that covers both customers and non-customers. But on their own, the ad servers can only track upper-funnel outcomes like click-through and viewthough, adding a new piece to the attribution puzzle, but not completing it.
Advertisers who invest in an integration of ad server and site data can develop a single map that charts all paid media exposure and their link to site outcomes. The most sophisticated advertisers include other "˜events’ in their maps, such as direct marketing e-mails and organic search referrals, and they anchor their attribution to more specific outcomes (like order size or lifetime value). With software (or a statistician!) to crunch the numbers, you can quickly assign credit where it is due.
That’s the current "˜state of the art’ when it comes to attribution modeling — miles ahead of "˜last-click’ attribution (which we still see in use all the time, by the way).
However, integration isn’t simple or inexpensive, and even the state of the art leaves four gaps that only an online panel can fill. Here are the four biggest gaps that we’ve seen that can be filled by using a panel for assigning attribution:
- The state of the art is largely blind to consumers’ exposure to "˜earned media’ (favorable blog posts, product reviews, viral videos) and to events like "˜search-throughs'; a panel-based view can incorporate these just like a banner ad.
- The state of the art doesn’t facilitate the collection of data about offline exposure and offline outcomes.Â Panel providers can survey consumers to learn about their exposure to ads on TV, radio, and print, and can enlist partners that identify offline transactions like grocery purchases —information that we can tie directly into the existing online-centric map.
- The state of the art doesn’t provide a view of macro trends in the industry that may impact outcomes (if sales are up for all competitors, it may mean that campaigns deserve less credit) or that may be triggered by media strategies (if a campaign’s "˜halo’ effect drives competitor conversions, then maybe it deserves some demerits); this market-wide perspective is a panel’s bread-and-butter.
By combining a panel’s broader and uninterrupted view of consumer behaviors with the rich volume of path-to-conversion data provided by ad serving and site analytics systems, marketers can finally aspire to solve the puzzle that attribution modeling has been for most advertisers.