“What’s the best way to measure marketing?” It’s the question I get asked more than any other, and the answer usually disappoints people. It depends on the question you’re actually trying to answer.
If your goal is to optimise a campaign in-flight, attribution can work. If you need to prove an activity caused an uplift, experiments are invaluable. If you want to know whether your investment is changing how people think about your brand, that’s brand tracking. And if you’re deciding where next quarter’s budget should go, that’s the job of marketing mix modelling.
Here’s the thing, though. If you’ve read anything about measurement in the last five years, none of the above is news. “Triangulate your methods” is now industry orthodoxy. Every platform, every consultancy, every conference panel says it. The interesting question isn’t whether marketers know no single tool can do all four jobs – it’s why so few organisations behave as if they believe it.
Walk into a marketing conference and everyone agrees you need multiple measurement methods. Walk into the budget meeting on Monday morning and it’s back to the last-click dashboard.
What last-click actually measures
Let me be precise here, because “attribution is broken” has become a lazy catch-cry that attacks the wrong target. Multi-touch attribution tried to credit the upper funnel. It failed mostly because the data (cross-device, walled gardens, privacy changes) made it impossible, not because the ambition was wrong. The version that actually runs most budget decisions is far cruder – last-click, or something close to it.
Take the most common example. A customer searches your brand, clicks a paid search ad and converts. The dashboard gives search the credit. Technically that’s correct, search captured the conversion. What it can’t explain is why that customer searched in the first place.
Perhaps they watched your television campaign during the AFL, heard your CEO on radio, saw the product on social, spoke to a colleague who already uses it, or simply grew more confident in the category over months. The search click was merely the moment demand became visible. Last-click measures where demand is captured, not how it was created, and it’s one reason search keeps absorbing such a large share of budgets. It’s measurable, accountable and extremely effective at converting existing intent.
But intent doesn’t appear by magic. Something has to create it first, and a lot of what creates it isn’t marketing at all.
The biggest blind spot is the economy
Consumers make purchasing decisions in a world where interest rates move, fuel prices swing, inflation reshapes household budgets, categories mature and competitors launch. Every one of those forces acts on demand before a single ad gets the chance to work.
Most measurement systems assume those conditions are constant. They aren’t. So campaigns get rewarded for running during favourable conditions and cut when confidence weakens – marketing praised for growth it didn’t create, and blamed for declines it couldn’t have prevented.
Strictly speaking, last-click never claimed to model interest rates – so call this an omission error, not an attribution one. It’s the omission that costs the most money.
63 per cent of BYD’s demand wasn’t media
We saw this firsthand with BYD. The original question sounded simple: which channels are driving test drives? Once we modelled the complete demand system – media, pricing, seasonality, competitors, macro conditions – the picture changed. Marketing directly influenced around 37 per cent of demand. The remaining 63 per cent was baseline: category growth, rising familiarity with EVs, brand momentum and macroeconomic tailwinds.
Now, the sharp objection, and if you’re technical you’re already forming it: how do you know that 63 per cent baseline isn’t swallowing the long-term effects of the media itself?
It’s the right question to ask of any MMM, including ours. The defence is in the build – brand strength modelled as its own driver, long carryover effects, and decomposition pressure-tested against evidence outside the model, not graded on its own homework. A baseline you can’t interrogate is a baseline you shouldn’t trust.
Some marketers hear “37 per cent” and think it diminishes marketing’s contribution. I’d argue the opposite. A CMO who walks into a board meeting claiming marketing drove 80 per cent of sales, only for the CFO to dismiss the number, has gained nothing. But a defensible 37 per cent, one that clearly separates marketing from macroeconomic tailwinds and can be validated against fuel prices, category growth and other external factors, gives marketing a credible claim on incremental demand. Smaller and believed beats bigger and doubted, every time.
Once we separated macro effects, we found efficiency differences of up to 20 times across the media mix. Some channels weren’t performing better because they were inherently more effective. They were simply active while fuel prices fell, consumer confidence improved or EV adoption accelerated. The economy created the demand; last-click attribution handed the credit to whichever channel happened to be there.
Remove that timing bias and the rankings change dramatically. That’s the difference between investing in the channels that got lucky, and the channels that genuinely drive growth.
Better decisions, not perfect answers
Measurement has never been about perfectly explaining the past. The CMOs I rate most highly are entirely comfortable with that, because business runs on incomplete information either way. Perfect was never the bar. The bar is simpler: does this evidence improve the next decision more than what you were using before?
The brands pulling ahead are the ones who stopped treating triangulation as a conference slide and started treating it as an operating discipline. Last-click for capture, experiments for causation, brand tracking for tomorrow’s demand, mix modelling for allocation – and each one checking the others’ work.
Everyone else can keep arguing about the perfect model. You’ve got decisions to make.
