Media mix modeling: why Mastercard and Uber revived it

September 8, 2026 · 5 min · Free, with full transcript

Was media mix modeling really obsolete, or did marketers pick fast flattering numbers over slow honest ones? The position here is that MMM never failed. Last-click attribution handed credit to paid search on a company's own brand name, and Uber's team admitted channels were claiming more revenue than the business made. Cheap cloud computing cut three-month models to overnight runs, and Apple's tracking opt-outs left attribution partly blind. You come away knowing where MMM fits against attribution, why Mastercard's anonymized aggregate transaction data suits it, how to treat SEMrush claims about brand search as a leading indicator, and a spend test you can run in two weeks.

Listen to the podcast

4 min

Chapters

Key takeaways

  • Pull your last-click report and check how much credit goes to paid search on your own brand name.
  • If branded search takes more than 20% of credit, cut that spend 30% for two weeks and watch whether sales actually drop.
  • Use media mix modeling for budget strategy and attribution for tactical creative and placement tweaks, not one instead of the other.
  • Treat vendor claims that brand search volume predicts MMM lift as a hint, and cross-check against your own sales curve.
  • Feed models aggregate purchase behavior rather than clicks, which is why anonymized transaction data has become useful.
Read the full transcript

Host:Welcome to the Leaders Insights Podcast. Today's episode, Media Mix Modeling, why MasterCard and Uber are betting on an old tool in a new era. Media Mix Modeling was supposed to be dead. It's the marketing equivalent of a fax machine. Why are MasterCard and Uber suddenly resurrecting it?

Expert:Because it was never dead. It was inconvenient. Media Mix Modeling, which is the statistical technique of estimating how each channel drives sales by looking at aggregate spend and results over time, gives you slow, honest answers. Digital attribution gave you fast flattering ones. The industry picked flattering.

Host:That's a strong accusation. Are you saying everyone knew attribution was lying and used it anyway?

Expert:I'm saying the incentives were rotten. Last click attribution, the model that credits whichever ad the customer touched right before buying hands the trophy to the channel standing closest to the finish line. That's almost always paid search on your own brand name. You're paying Google to take credit for demand your TV ad created.

Host:Come on, search still drove the conversion.

Expert:Did it? If someone Googles Uber and clicks the ad, the ad didn't create that intent. The brand did. Uber's own team admitted they were double counting. Everyone said their channels drove more revenue than the company actually made. When your math claims 130% of sales came from your ads, the math is the problem, not modesty.

Host:So MMM is just the grown up in the room. Why did it fall out of fashion in the first place?

Expert:It was genuinely painful. Old school Media Mix Modeling took a team of statisticians three months to deliver a report that was already stale. You'd learn what worked last spring sometime around Christmas. Digital attribution refreshed every morning. Speed beat accuracy and for a decade, nobody got fired for it.

Host:What changed? Computers were fast in 2022.

Expert:Two things. Cheap cloud computing means a model that took three months now runs overnight and privacy rules gutted the tracking that attribution depended on. When Apple let people block the identifier that followed them between apps, roughly a quarter of iPhone users opted out immediately. And attribution went partly blind. Suddenly the precise tool couldn't see and the old tool that never needed to track individuals looked brilliant.

Host:MasterCard's a payments company. What's their angle here?

Expert:They sit on aggregate spending data across millions of transactions, anonymized, no individual tracking. That's exactly the Fuel Media Mix Modeling wants. They can feed a model actual purchase behavior instead of clicks. MasterCard isn't nostalgic for an old tool. They own the ingredient that finally makes it work.

Host:SEMrush data suggests brand search volume is a leading indicator of MMM lift. You buy that?

Expert:I'd use it as a hint, not gospel. SEMrush sells search and analytics software, so their whole business depends on you believing search metrics predict everything. Cross-check it against your actual sales curve before you build a budget on it. Directionally, brand searches rising after a TV flight is real. The precise coefficient they'd sell you, less so.

Host:Here's the uncomfortable one. If MMM is so good, why not fire the attribution team entirely?

Expert:Because MMM tells you the what and misses the who. Media Mix Modeling says TV drove $2 million last quarter. It can't tell you which specific ad to swap out on Tuesday. Attribution's terrible at strategy, but fine for tactical tweaks. You need both. Firing attribution to worship MMM is just swapping one religion for another, and I've watched CMOs do exactly that and blow their next quarter.

Host:So the honest answer is, it depends, which is what every consultant says to keep the invoice open.

Expert:The honest answer is measure the machine, not the moment. Attribution obsesses over the last touch, because it's easy to see. That's like judging a football match by who tapped the ball into an empty net and ignoring everyone who ran it up the pitch.

Host:Fine. One thing a listener does Monday morning.

Expert:Pull your last-click report and find how much credit goes to search on your own brand name. If it's over 20%, you're paying to take credit for demand you already created. Cut that spend 30% for two weeks and watch whether sales actually drop. If they don't, you just found free money.

Host:Sources for today's episode. DigiDay, Adweek, Semrush, Vendor, SEO Analytics Tools. That's it. The CMO Decision Tools are live at MBA-training.com.

Read and go deeper