Marketing & Advertising · Incrementality

Verify combined platform ad results with independent lift testing

In short Before shifting budget to a combined retail media platform, have an independent party measure incrementality against your own data, because platforms grading their own ads can't be trusted by default.

Digiday reports that advertisers are asking what happens to costs, data and measurement now that DoorDash and Instacart are joining forces.

Costs and data are fair questions. Measurement is the one I'd press hardest.

A bigger combined platform means one seller reporting sales results on its own ads. That's grading your own homework. The dashboard can look great and still shouldn't be trusted by default.

Platform-attributed sales count every buyer who saw or touched an ad. Plenty of them would have bought anyway. Incrementality tells you which ones wouldn't.

So before you move more budget to the combined offering, get an independent party measuring lift against your own data, like your analytics. Run a holdout test. Put that number next to what the platform reports.

If they match, good. If they don't, you've learned how much you were paying for sales you'd already won.

Responding to Here are the questions on advertisers’ minds after the DoorDash-Instacart deal, Digiday

Originally posted on LinkedIn →

Topics Incrementality · Retail media · Marketing measurement · Advertising

Jeff Greenfield

About the author

Jeff Greenfield

Builds marketing measurement platforms, from multi-touch attribution to incrementality

Jeff Greenfield has spent two decades building marketing programs tied to measurable value. In 2008, Jeff built one of the first multi-touch attribution platforms, which earned MRC Accreditation. In 2022, Jeff launched Provalytics, which measures incremental impact across channels without cookies, tags, or user-level data.

More from Jeff →

Work with Jeff

Send a message through the Index. Jeff replies directly.

Your message is emailed to Jeff. We don't share your address with anyone else.