Behind the Numbers: an EMARKETER Podcast

The episode surfaces two genuinely useful data points about iROAS variance, but the remaining ~10 minutes are Cannes small talk, grocery-shopping anecdotes, and high-level platitudes about measurement maturity. Insight-per-minute ratio is low for a 13-minute runtime.

“over 80 of campaigns can shift from positive to negative IROAS, depending on methodology”

“measurement isn't a report to put on the fridge. Measurement should be used to inform your next move”

The 80%-flip finding is a genuinely striking framing, but the broader narrative - retail media needs incrementality standards, measurement is evolving, retailers and brands have aligned interests - is well-worn 2024 industry discourse recycled in a Cannes setting.

“there's one way to slice an avocado, but there are a lot of ways to peel an orange. And I put retail media and measurement more in the orange camp”

“we've been long spoiled by the very tight attribution loop of one-to-one”

Liz Roche is a genuine VP-level measurement practitioner at a major retail media network and demonstrates real technical fluency around control group design; however, Albertsons is the episode sponsor, which constrains candor and positions this partly as a marketing exercise.

“If you're using a match control, what are some of the factors that you're matching on? If you're using synthetic control, if you're using RCT, understanding that methodology I think is really important”

“I be honest I asked the team to double check the math”

The episode contains exactly two concrete numbers (6.5x variance, >80% of campaigns can flip sign) drawn from a named third-party study with Ovative, but almost every other claim is left abstract - no named brands, no campaign budgets, no timelines, no breakdown of which methodological choices drive the variance.

“results can vary 6 I think the biggest shock to me though was that over 80 of campaigns can shift from positive to negative IROAS, depending on methodology”

“research that Albertsons just put together in partnership with Ovative on the topic of incrementality or IROAS”

The host asks a few functionally useful questions (methodology shock, advertiser response, governance vs. performance) but never presses on the critical details - what specific methodological choices cause the 6.5x swing, or why Albertsons' own approach should be trusted - and explicitly softens potentially pointed observations before the guest can address them.

“I think it's fair to say acting in good faith”

“Did anything about the findings surprise you or shock you? Or did it just sort of validate something that you had had sort of a hunch about?”