Marketing measurement

From ROAS to profit curves: a finance-first method for cross-media budgets

In 2023, while leading business finance for Google in France, Italy, Spain and Portugal, I described a simple financial model for one recurring question: how much should a company invest in marketing, without over- or under-investing? The idea, published on Think with Google, was to stop judging spend by return on ad spend alone and to place each advertiser on a profit curve.

This article recaps that idea and shows how we extend it at 40folds: from a single channel to a full cross-media measurement method, with MMM-style models used as an input to estimate the saturation curve of every channel.

The original case: growing volume, shrinking profit

The starting point was an omnichannel retailer active in several markets. Its Search account was fully optimised and automated, with clear cost-of-sale targets agreed with finance for each channel. Chasing click share against competitors, it had kept raising spend year after year — and total profit had started to fall.

We combined observed results — historical data, click share — with business inputs such as volume elasticity and margin contribution to compute the “true” return generated. Put together, these give a profit curve: incremental net profit as a function of investment.

The retailer still cleared its minimum hurdle rate, so the channel looked healthy. But it sat well to the right of the maximum-profit point: over-invested. Each additional euro bought volume at a loss, eroding total profit. Pulling spend back towards the maximum-profit point would give up some revenue but deliver more net profit and a much higher ROI.

The best ROI is rarely the best profit. The question is not “is this channel efficient?” but “where on the curve should we be?”

Why the curve matters to a CFO

A profit curve turns a marketing debate into a finance decision. It shows three points everyone can agree on: the hurdle rate (the minimum acceptable return), the maximum-profit point (where the marginal euro stops paying back), and the zone in between where the company trades ROI for volume. Choosing a point becomes an explicit strategic choice, not a negotiation between departments.

The limit: one channel at a time

Applied channel by channel, the method has a blind spot. Channels share the same customers and the same demand: TV lifts search, social feeds brand queries, promotions borrow from future sales. Optimising each curve in isolation double-counts impact and ignores that the marginal euro could earn more elsewhere.

Extending it to cross-media measurement

To go from one curve to a whole portfolio, we need one response curve per channel, estimated consistently and on the same baseline. This is where marketing mix models are useful — not as the final answer, but as an input.

• Estimate each channel’s saturation curve and carry-over (adstock) with MMM-style models, controlling for price, seasonality, distribution, macro conditions and competition
• Run many model variants rather than one, so each curve comes with a credible range instead of a single line
• Convert volume response into net profit using margin contribution, volume elasticity and cost of sales — the finance layer from the original model
• Allocate the budget across channels where marginal profit is equal, subject to the hurdle rate and any volume objective
• Penalise moves whose gain is large on average but highly uncertain, so recommendations stay risk-adjusted

The result is a cross-media profit frontier: for any total budget, the best split between channels, and for any split, the expected profit with its uncertainty. The single-channel question “are we over-invested in Search?” becomes the portfolio question “where does the next euro create the most profit, across all channels?”

Validating the curves

Saturation curves from observational data are fragile. We cross-check them against geo or holdout experiments, platform lift studies and attribution data where available, and keep only the model families that agree with these tests. Where evidence is thin, the curve stays wide — and the recommendation stays cautious.

From measurement to a shared decision

What made the original model useful was not the maths, but the conversation it enabled between marketing and finance. Extending it to the full media mix keeps that spirit: one consistent view of how each channel saturates, expressed in profit and risk — the language every member of the executive committee already speaks.

Source: R. Rusu, “Un modèle financier pour garantir le niveau d’investissement marketing optimal”, Think with Google, July 2023 — Read the article