The brands that win long term aren’t just outspending the rest. They’re out-learning them, writes Sophie Murphy, marketing science partner at Mutinex.
Marketing isn’t magic, it’s more like science. But there is magic in testing. The best brands approach it like scientists: forming hypotheses, running controlled experiments, learning from the data, and refining as they go. It’s not guesswork, it’s a disciplined approach to creating measurable change.
Not to be dramatic (although we know I can be), but testing might just be the unsung hero of modern marketing.
There’s often this idea that testing slows things down. That it’s what you do after a campaign flops, not before. But it’s actually the opposite. Testing helps you learn faster, spend more efficiently, and scale with confidence. The brands doing this well aren’t necessarily spending the most, they’re just making their dollars work harder by using data to ask the right questions upfront, not after the fact.
So, what actually makes a good test?
It’s not about throwing budget at a new channel and hoping for fireworks. A proper test-and-learn starts with a clear and falsifiable hypothesis. A simple, testable prediction about the effect of one variable on another. Think: “If we consolidate investment from a 30-second video format to a 15-second format, we expect to see higher revenue generation.” Simplicity is key, because the clearer the test, the clearer the conclusion.
Then comes the validation. The bit where you check if your prediction actually held up. This is where tools like MMM come into play, helping you confirm whether the outcome really happened. And good models follow proper governance, so you can understand the odds of the outcome. No smoke, no mirrors, just a reality check.
And here’s the kicker, being wrong is not a failure, it’s a feature. Good tests aren’t about proving yourself right, they’re about learning something useful. Disproving your hypothesis? That’s gold. It means you’re getting closer to the truth, and that’s the whole point. Or, as Henry Ford put it, “Failure is simply the opportunity to begin again, this time more intelligently.”
And notice, a good test doesn’t need to involve more budget. Of course, it can, but the biggest learnings often come from the smaller, smarter shifts. Testing different formats lengths, shifting flighting, re-distributing spend across publishers, trialling geo performance. These micro-moves can uncover insights that punch well above their weight.
And we couldn’t talk about this without acknowledging the obvious: creative is half the picture. Nielsen suggests nearly 50 per cent of media’s effectiveness comes from creative quality. Just like media, it’s about small tests, smart learnings, steady improvements. That’s the game. You won’t find the perfect balance straight away, but testing shows you what’s worth repeating, and what’s not.
And if you have a good MMM to co-pilot these tests, ideally one that’s Bayesian and time-varying at a monthly cadence, it becomes much more than a backward-looking reporting tool. It’s a forward-looking simulator. A Bayesian model allows you to combine what you already know about marketing (like diminishing returns, lag effects, or channel synergies) with new data, constantly refining its predictions. And because it’s time-varying, it adapts to how marketing performance changes over time, across months, seasons, or market conditions. This is essential for testing, where timing and context can shift the impact of your spend.
Together, this means you’re not testing in the dark, you’re testing with a spotlight. You can model different future scenarios, understand likely outcomes, and weigh trade-offs before anything goes live. No scrambling after the fact, just clear guidance on where to test, when to test, and what’s worth testing next.
Ultimately, testing isn’t about being risk-averse, it’s about being risk-aware. It’s a more informed way of working. One that says: let’s understand what works before we double down.
So, your next move doesn’t need to be bigger, it just needs to be data backed. What can you test to make sure it is? Because the brands that win long term aren’t just outspending the rest. They’re out-learning them.
