How did last quarter’s split actually get decided?
Think back to the last time your channel budget got reallocated. It probably started with last quarter’s numbers as the baseline, drifted toward whichever dashboard looked best in the room, and settled wherever whoever argued hardest wanted it to land.
If that sounds familiar, you’re in good company. A lot of budgets get set this way, including good ones, and it says nothing about anyone’s judgement. The people in that room are doing their best with what they’ve been handed. The tooling was never built to answer the question a budget meeting needs answered, and a split defended on confidence rather than evidence gets re-argued the same way next quarter, by whoever is loudest that time around.
Why the attribution report on your desk can’t settle it
Here’s the mismatch. Last-click and platform-reported attribution both describe what already happened: which touchpoint, or which platform, got credit for a conversion that had already closed. A budget meeting is asking something else, which is where the next dollar does the most good. Describing the past accurately doesn’t answer a question about what to do next.
That’s why a channel that reliably closes deals and a channel that reliably opens them can look identical on the report, or worse, make the closer look like the only thing working. Last-click credits the final touch, so the channel that made the buyer ready to buy earlier gets nothing.
There’s a second reason the report falls short. A meaningful share of iPhone users decline app tracking, and browser privacy limits do similar work on desktop, so part of the customer journey never reaches the report at all. Whatever the report says, it’s built on partial data.
Same month, three methods, three possible answers
Take one month of channel spend and look at it through three common lenses. Each one answers a different question.
| Method | What it credits | What it can miss |
|---|---|---|
| Last-click | The last touchpoint before the sale | Anything earlier that made the buyer ready to buy |
| Platform-reported | Whatever each platform’s own tracking claims as a conversion | Sales that several platforms each claim, and journeys that cross devices or channels |
| Marketing mix modelling | Each channel’s share of the overall result, estimated from spend and sales over time, with seasonality, pricing and competitor activity accounted for | The path of any one customer, and it needs enough history to be reliable |
Same spend, same sales, and each lens can point to a different channel as the one that’s working. A channel that looks like the hero under last-click may look ordinary once a model accounts for what would have sold anyway. A channel that looks expensive in platform reports may be doing work earlier in the buying process that the other two lenses weren’t built to see.
That’s the question most budget meetings end on without naming it: which of these three answers would you take to your CFO?
None of them is wrong. They answer different questions, and the one a budget decision rests on is the third: what actually moved the result.
So what is marketing mix modeling, in plain terms?
Marketing mix modelling (MMM) takes spend by channel over time, along with the outside factors that also move sales, such as seasonality, pricing and competitor activity, and estimates statistically how much each channel contributed to the outcome. It works from your own spend and sales history, and where that history is thin it can lean on assumptions from past tests or industry benchmarks. What comes out is a contribution estimate for each channel with a stated range, because a single precise-looking number would claim more than the method can know.
What a model measures is what moved the outcome once everything else in the market is accounted for. That’s the question attribution wasn’t built to answer. Attribution is good at the other one, who got the credit.
What it takes to run one
Here’s what you’re signing up for, because being clear about the cost is what makes the recommendation worth trusting.
A model needs enough history, and enough variation in how you’ve split spend, to learn anything useful. A brand that has run roughly the same channel mix for years gives a model little to compare, so there’s little for it to learn from.
A first model takes calendar time, and it won’t hand you one certain number to quote without qualification. It gives you a range and a confidence level. That’s a better foundation for a decision than a guess dressed up as certainty, and a range still leaves room for judgement. Anyone promising an exact figure from a first pass is promising more precision than the method delivers.
Where we’d start if you wanted to move on evidence
A budget mix audit looks at your current channel split against what your own spend history can tell you. It answers one question: is the gap between methods big enough in your account to be worth acting on, before you spend anything on a full model.
We’re willing to say the gap might not be worth chasing yet, and that’s what makes it worth listening to us on the accounts where we say it is.
Get a budget mix audit.
