Tracking users across sites and across touchpoints is increasingly difficult. Between regulatory constraints and the increasing blocking and rapid expiration of third-party cookies, what is the marketer to do when it comes to determining the impact of their multi-channel (omnichannel!) marketing‽

At our February 2021 (virtual) meetup, data scientist Dr. Joe Sutherland from Search Discovery described both the challenges (and inherent shortcomings) of “traditional” multi-touch attribution and then put forth a radical, but, in many ways, not at all new alternative: randomized controlled trials (RCTs). In a nutshell, it’s like an A/B test…but using randomly assigned geographies (zip codes, DMAs, etc.) to “split” the media. This includes everything from a “basic blocking”  design (straight up “on in some, off in others”) to “stepped wedge” designs (every geo gets some media, but the level and timing of the media varies).

This approach provides truly causal results, which is something that media mix models and traditional multi-touch attribution simply could never do! Plus…it doesn’t rely on user-level tracking (it’s based on where investment is made and the results of those investments). And there was a chihuahua muffin (it’s a machine learning thing):

The slides from the presentation are available, or, if video is your jam, then you can live/re-live the experience that way:

Several books and podcasts were recommended during Joe’s presentation and in the chat:

• (Book) Subprime Attention Crisis: Advertising and the Time Bomb at the Heart of the Internet, by Tim Hwang
• (Podcast) Freakonomics Radio episodes 440 (traditional advertising) and 441 (digital advertising)
• (Podcast) Digital Analytics Power Hour podcast episode with Dr. Joe Sutherland on this very topic
• (Podcast) Digital Analytics Power Hour podcast with Tim Hwang