Latest / MBA Training Marketing / How Procter & Gamble rebuilt its media mix modeling capability and what it actually changed
Transcript
- 0:00Welcome to the Leaders Insights podcast.
- 0:08Today's episode how Proctor & Gamble rebuilt its media mix modeling capability and what
- 0:13it actually changed.
- 0:14Proctor & Gamble spent the better part of a decade tearing down the exact measurement
- 0:18system they've now rebuilt.
- 0:20That's not a pivot.
- 0:21That's an apology.
- 0:22What did they get wrong?
- 0:24They believed the platforms.
- 0:25That's the whole story.
- 0:26Facebook, Google, the ad tech vendors.
- 0:29They all handed P&G attribution dashboards, showing every dollar tracing neatly to a click.
- 0:35And P&G decided that clean-looking data must be true data.
- 0:39But attribution tracing a sale back to the ad that caused it sounds like exactly what
- 0:44a marketer should want.
- 0:45It sounds like it.
- 0:46Here's the con.
- 0:48Every platform grades its own homework.
- 0:50When Google claims a conversion and Facebook claims the same conversion and your email tool
- 0:55claims it too, you've now paid for that one sale three times over in your internal
- 1:00reporting.
- 1:01P&G was summing those numbers and finding they'd attributed more sales than they
- 1:05actually made.
- 1:11They noticed around 2019 when Mark Pritchard's team cut over $200 million in digital spend
- 1:17and sales didn't move.
- 1:19Read that again.
- 1:20They turned off the ads the dashboard swore were driving growth and nothing happened.
- 1:25That's your tell.
- 1:27If you can remove the input and the output holds, the input was never doing the work.
- 1:31So they went back to media mix modeling.
- 1:34Explain that without putting me to sleep.
- 1:37Media mix modeling.
- 1:38Here we, the statistical approach that looks at total sales and works backwards to estimate
- 1:42what each channel actually contributed using economy-wide data instead of platform-reported
- 1:49clicks.
- 1:50It's old.
- 1:51It's from the 1960s.
- 1:52And that's precisely why the ad tech crowd spent years calling it a dinosaur.
- 1:57They weren't entirely wrong, were they?
- 1:59The old models took months and told you nothing timely.
- 2:02True.
- 2:03And I won't defend the 1990s version.
- 2:05A quarterly report telling you last spring's TV spend worked is useless when you're
- 2:10allocating budget today.
- 2:12What P&G rebuilt is faster.
- 2:14Models refreshed roughly every two weeks.
- 2:17And in-house rather than outsourced to a consultancy charging by the slide.
- 2:21In-house is the interesting word.
- 2:23Why does that matter?
- 2:24Because when a vendor builds your model, the model tends to flatter the vendor's product.
- 2:29Nielsen's data.
- 2:30When Nielsen's selling you the study, you can guess which channel comes out looking
- 2:34essential.
- 2:35P&G brought the statisticians onto their own payroll.
- 2:39So the answer wasn't for sale.
- 2:41Here's the uncomfortable one.
- 2:43Can't we build it ourselves just a luxury for a company spending $8 billion a year on
- 2:47advertising?
- 2:48What does a normal CMO do?
- 2:50Fair hit and mostly yes.
- 2:52But the transferable lesson costs nothing.
- 2:55Run a holdout test, pick a region, shut off a channel completely, and watch what happens
- 3:00to sales versus a matched region where you left it on.
- 3:04That's the one experiment that can't lie to you.
- 3:07And it works whether you spend $8 billion or $800,000.
- 3:10You're telling marketers to deliberately stop advertising in a market.
- 3:14Their bosses will love that.
- 3:16Their bosses will love finding out that 30% of the search budget was buying customers who'd
- 3:21already decided to buy.
- 3:23Which is roughly what these geo tests keep surfacing.
- 3:27The paid ad just took credit for the sale on its way out the door.
- 3:30Turning it off is how you separate the ads that persuade from the ads that photobomb.
- 3:35So what actually changed at P&G beyond the org chart?
- 3:38They stopped optimizing to the click and started optimizing to the model.
- 3:42Concretely, they shifted spend back toward brand building.
- 3:46The stuff no attribution tool can measure because there's no click to count.
- 3:51And their reported marketing efficiency by their own 2025 disclosures improved without raising
- 3:57the budget.
- 3:58Flag it.
- 3:59That's P&G's own number, not an auditor's.
- 4:01Give me the one thing a listener does Monday morning.
- 4:04Pick your single biggest paid channel.
- 4:06Turn it off in one clean geographic market for four weeks.
- 4:10If sales don't drop, you just found your budget for next quarter.
- 4:14And the dashboard that told you otherwise was writing itself a very nice performance review.
- 4:19That's it.
- 4:20The CMO Decision Tools are live at MBA-training.com.