Stop letting your ad platforms grade their own homework
Attribution is not an analytics problem. It is a governance problem. Why one blunt, ungameable number should anchor marketing’s accountability to the CFO, and how incrementality testing earns back the right to optimize channels.
Every major ad platform reports performance using its own attribution logic, on its own data, to a customer whose budget it is paid from. Taken together, they routinely claim more revenue than the business earned. Most marketing organizations know this and allocate budget on those numbers anyway.
Attribution is a governance failure dressed up as an analytics problem. The fix is not a better model. It is a single, ungameable measure of marketing’s contribution, agreed with finance, and a testing discipline that makes channel claims earn their way into budget decisions.
The arithmetic of self-reported success
Hypothetical illustration. Overlapping credit routinely exceeds actual revenue by a wide margin.
Two structural biases drive the gap. Credit is duplicated across every touch a buyer had. And platform defaults drift toward audiences who were already going to buy, which produces the best-looking returns exactly where incremental impact is lowest. Budget follows the flattering number, and the business funds activity that changes nothing.
Anchor on blended MER, agreed with finance
Blended marketing efficiency ratio, total revenue over total marketing spend, is deliberately blunt. It cannot tell you which ad worked. It can tell you, without debate, whether the marketing system as a whole is paying for itself, and it is the one number marketing and the CFO can both sign.
Its value as a management instrument shows up under stress. At Flax Labs, a restructure pushed blended MER as low as 1.8x. Because the whole team could see the number, every subsequent decision, from role design to which clients tested new tooling, was judged against it. Within a quarter it was back at 5.6x. Across client accounts, MER ranged from 2.8x to 13.7x, which is the point: there is no universal good number, only your floor, your trend and whether your decisions move it.
Pair it with a margin floor
MER without margin is vanity at a higher altitude. A strong ratio on a thin-margin product can still destroy value once cost of goods, fulfilment, returns and discounts are counted. The governance model I use sets a minimum MER with finance, derived from contribution margin, and treats anything below it as growth the company is paying for. Deliberate dips, such as market entry or brand investment, are budgeted as investment with an expected duration, not discovered in a quarterly review.
Make channels earn their claims
Channel data still matters. It just moves from fact to hypothesis.
| Method | Question it answers | Where it earns its keep |
|---|---|---|
| Holdout test | What happens to this audience without the spend? | Person- or account-level targeting |
| Geo-lift | What happens to regional revenue when spend pauses? | Broad-reach and offline-influenced channels |
| Marketing mix model | What is each channel’s marginal return, and where does it saturate? | Portfolio allocation, calibrated by real tests |
At BuildDirect I ring-fenced 10% of a paid budget that grew from $3.5M to nearly $9M specifically for testing. Protecting that line mattered more than any single test, because testing is the first thing cut when results soften, which is exactly when the business most needs to learn. ROAS rose 230% over that period.
How I set the floor with finance
The minimum acceptable MER should be derived, not negotiated. Start with contribution margin after cost of goods, fulfilment, returns and payment costs. Decide what share of that margin the business is willing to reinvest in acquisition at its current stage. The ratio that results is the floor. A business with high margins and an aggressive growth mandate will set a lower floor than a thin-margin business optimizing for cash. Writing the derivation down matters as much as the number: when conditions change, the board can see which assumption moved.
Core versus investment spend
Not all spend should be held to the same standard in the same quarter. Market entry, new product launches and brand programs are investments with delayed payback. I separate them in the plan and in reporting, with an expected payback window agreed upfront. Core spend is held to the floor every week. Investment spend is held to its window. This prevents two common failures: cutting long-horizon programs because they depress this month’s ratio, and hiding underperforming core spend inside an “investment” label.
Where blended MER misleads
No single metric is complete, and leadership should know the limits of this one. Blended MER can flatter a business that is harvesting an existing brand while under-investing in future demand. It can obscure channel saturation if one strong channel masks another that has stopped returning. And in businesses with long sales cycles, revenue lags spend by months, so the ratio should be read as a trailing trend rather than a weekly verdict. The remedy is not a more complex number. It is pairing MER with incrementality tests, a margin floor and a separate view of investment spend.
The governance rhythm
Weekly: blended MER and contribution margin against the floor, reviewed by marketing and finance together. Monthly: results of any incrementality tests completed, and the budget moves they justify. Quarterly: the mix between core and investment spend, the payback status of each investment and the recalibration of any marketing mix model. Annually: the floor itself, recalculated from current margins and strategy. This rhythm turns marketing measurement from a quarterly argument into a standing, shared view of capital efficiency.
Questions for the board
- What is our blended MER, what is our agreed floor, and how was the floor derived?
- How much of last year’s budget moved on the basis of an incrementality test rather than platform-reported results?
- What share of spend is classified as investment, and when is it expected to pay back?
The takeaway
Stop letting the platforms that sell you media tell you whether it worked. Run marketing on one honest number agreed with finance, protect the budget that tests channel claims, and let attribution generate hypotheses rather than decisions.