Half Your Ad Budget Is Going to Waste? How an FMCG Giant Used AI to Finally Do the Math

The other night, I had dinner with a friend who works in FMCG (i.e., fast-moving consumer goods).

He runs marketing for several product lines at his company, with annual budgets in the tens of millions of dollars. Somewhere between dishes, he sighed: every year at planning season, his boss asks him the same question—"Last year's money. Was it well spent?"

He told me he can't answer it.

Which channels pull their weight? Which campaigns are just burning money? Where should next year's budget go, and where should it come from? He has a feeling. He doesn't have numbers.

And he's not alone. That century-old line from the advertising world still rings true today: I know half of my advertising is wasted—I just don't know which half.

Then the pandemic years scrambled how people buy. Trying a new brand got easier, and loyal customers walk away without a second thought. E-commerce and direct-to-consumer (DTC) channels keep multiplying, so new brands surface more easily while loyalty to the old ones thins out. For FMCG marketers, the budget has never been harder to spend well.

What Is Marketing Mix Modeling?

Let me walk you through it.

Every marketing dollar you spend—on TV, in-feed ads, e-commerce mega-sale events, offline channels—eventually flows into one big pool called sales. Marketing Mix Modeling (MMM), to use the industry's name for it, takes that pool apart and does the accounting: how much sales did each stream contribute, and did it earn its keep?

Put simply, it puts a scale under your marketing.

Sounds great. But once you try to actually use it, the sticking point isn't the model. It's the data.

I came across a textbook example.

A Fortune 500 FMCG giant: dozens of brands, sold in markets around the world, a sprawling lineup of online and offline channels, and more than $100 million a year in marketing spend. For years, its marketing data lived in silos—traditional media in one pile, digital campaigns in another, consumer surveys in a third, attitude research in a fourth, sales data in a fifth. Different formats, different definitions; even the timelines wouldn't line up.

Unified measurement? The pieces wouldn't fit.

So every product line fought its own fight, and budget allocation ran on experience and nerve. Plenty of money spent. Nobody could say how much of it was wasted.

Then a data analytics firm, WNS Analytics, took the job. The interesting part: they didn't open with a showcase of algorithms. They started with the least sexy task of all—cleaning the data.

Step one: integrate. Gather the scattered data—traditional and digital media, surveys, attitude research, sales records—align it to one measurement standard, and pour it into a single measurement platform. No foundation, no building.

Step two: model. Using AI and machine learning, they built a marketing mix model for each brand individually, working out the true effect of every marketing lever. One detail I find especially important: the models get regular health checks. Markets shift, consumers shift; a model that was accurate last year will quietly start lying to you if you leave it alone. The industry term is "drift."

Step three: distill. The insights were organized into an omnichannel playbook—how consumers respond to campaigns, which plays work—so the lessons become reusable. Measurement isn't a one-off project; it's a loop of continuous improvement.

Step four: ship tools. WNS equipped the marketing team with a simulator and an optimizer, simple enough for business users to run themselves. Want to move budget around? Test-drive the scenario in the sandbox first and watch what happens to sales. The results plug straight into media partners' tools too, so media planning no longer means shuttling data back and forth.

Time to Do the Math

This system monitors more than $100 million in marketing spend.

On top of that base, total sales rose by more than 3%, and profit grew by more than 5%.

Honestly, when I read that not a single dollar had been added to the budget, I did a double take. No new money spent—so where did the extra sales and profit come from? Carved out of the share that was being wasted all along.

3% sounds unremarkable, right? For a company spending nine figures on marketing every year, it's found money.

No new investment. Just moving the money that was already there. That is what measurement buys you.

Worth more than the 3%, though, are two other things.

One is the unified view. Campaign performance for every brand and every market sits on the same table—which lines are strong and which are soft, all at a glance. Even the internal squabbling dies down.

The other is self-service. Once the simulator and optimizer spread across the whole organization, budget planning stopped being "queue up at year-end for the analysts' report" and became something marketers run themselves, any time. Decision-making moves at a completely different speed.

There's an old line in management: you can't manage what you can't measure.

Marketing more than anything. Spending money and hearing no echo back isn't investing—it's wishing. What can be measured can be optimized.

Back to That Dinner

As the meal was winding down, he asked me: should a company like ours get a system like this too?

I said: don't rush to the model. Sort your data first. If you can't even get the numbers from your various channels onto one sheet, what exactly are you planning to model?

He nodded, pulled out his phone, and started taking notes.

And a wish for you, too: may every marketing dollar you spend be a dollar you can trace.