An owner usually sees marketing as one line in the budget and two numbers in the board pack: a blended customer acquisition cost and the return the ad platforms report. Both are easy to produce, and both say little about whether the next million is well spent. This guide is for PE owners, operating partners and CFOs who want to check a portfolio company's marketing efficiency on a basis that holds up: whether the budget is the right size, what the next customer costs, how to fund a new country, and how to compare several companies on the same terms.
Blended CAC and platform ROAS describe the average of what was spent. The question an owner needs answered is at the margin: what the next part of the budget returns in each channel, and whether the total is above or below the level where extra spend stops paying for itself. A marketing mix model estimates that from the company's own spend and sales over time, digital and offline together, with a range on every number. Ask for the marginal return per channel against the return you require, marginal CAC against lifetime value, and forecasts that were stored before the actual sales arrived and checked against them.
Three habits make marketing look better or worse than it is.
None of this means the marketing team is doing a poor job. The numbers it is given were built to run campaigns, and an owner needs numbers built to size an investment.
Every channel has a response curve: the sales it brings at each level of spend, rising steeply at first and flattening as the channel saturates. The slope at today's spend is the marginal return, what one more unit of spend earns. Add the channels up and you get the curve for the whole budget.
The owner's question then has an answer in the owner's terms. Set the return the business requires on the last part of the budget. Where the marginal return is above it, more spend pays. Where it is below, the budget is larger than it should be, even if the average return still looks good. A scenario planner shows the same thing as forecasts: the current budget, a cut and an increase, each with a revenue forecast and a range. How to set the right budget level and board-ready scenarios go into both.
Ask for the range
A single number for the effect of a channel invites false confidence. A model that reports a range shows where the evidence is strong and where it is thin. A wide range on a large channel is a finding in itself: it is where a structured test is worth its cost.
CAC against lifetime value is the right test, often applied to the wrong CAC. The blended CAC divides total spend by total new customers, so it mixes the cheap customers who would have come anyway with the expensive ones won at the edge of each channel. The budget decision is about the edge.
Compare the marginal CAC, the cost of the next customer in each channel, with lifetime value over the payback period the owner accepts. When marginal CAC is below LTV over that period, more spend creates value. When it is above, the company is buying customers that do not pay back, whatever the blended figure says. For subscription businesses and software, see subscription and DTC and software and SaaS.
A new market has no history, so its first budget cannot come from its own data. What the model has learned in the home market, how fast each channel saturates and how long its effect lasts, is carried over and scaled to the new market's size as the starting point. The first months of spend then work as the test: the new market's own sales correct the starting point as they come in, and the range on the recommendation narrows.
For an owner, this turns market entry into a plan with checkpoints: a starting budget with a stated uncertainty, and a point at which the evidence says whether to push on, hold or pull back.
Comparing marketing efficiency across companies only works when the numbers are built the same way. That means one data structure for every company: market first, then product and segment, then the media, online and offline together. It means one model per market, since markets behave too differently to share one. And it means the same outputs from each: marginal return per channel with a range, marginal CAC, and a record of forecasts checked against actual sales.
Built that way, the comparison is fair across companies of different size and channel mix. A company whose marginal returns sit well above the bar has room to grow its budget. One whose marginal returns sit below it has money to move or to save. The record of checked forecasts tells the owner how much weight each company's model deserves.
The model needs weekly spend per channel and sales over time, two years at the least, plus the dates of price changes, promotions and other events. Digital spend comes through the ad platforms' APIs. TV, radio, print and out-of-home come from media plans and spot reports, usually from the media agency. The data guide lists it all.
Odins is a managed marketing mix modelling platform: we collect and structure the data, build and maintain the model, and review every recommendation before it reaches the company. A first model is ready 4 to 6 weeks from kickoff, data collection and structuring included; once the data is in, the model itself takes two to three weeks. Vendors who quote shorter timelines usually count from the day your data is already collected and prepared, and often do not offer data collection at all. After that, data flows in daily, the model is retrained monthly with a reviewed recommendation, and scenarios and forecasts can be run any time, in the platform or from Claude through the Odins MCP.
Because it is built on the company's own data and takes a few weeks, this is a tool for companies you own: value creation and board reporting during the holding period, not a check before a deal. Odins fits companies with a marketing budget from EUR 1M to several hundred million a year. What a model costs, and how to compare quotes over two years, is in how much marketing mix modelling costs.
Ask for three things the usual reports do not give: the marginal return per channel against the return you require, marginal CAC against lifetime value, and forecasts that were stored before the sales arrived and checked against them. A marketing mix model built on the company's own spend and sales provides all three, with a range on each number.
On the margin. Compare the cost of the next customer in each channel with lifetime value over the payback period you accept, not the blended CAC. The blended figure mixes customers who would have come anyway with the expensive ones at the edge of each channel. See how CFOs use Odins.
Start from what the home market's model has learned about each channel, scaled to the new market's size, and treat the first months as the test. The new market's own sales then correct the starting point, and the range on the recommendation narrows as the evidence comes in.
Yes, when every company is modelled on the same structure: one data hierarchy, one model per market, and the same outputs with ranges. Then marginal return and marginal CAC compare fairly across companies of different size and channel mix, and the record of checked forecasts shows how much weight each model deserves.
4 to 6 weeks from kickoff, data collection and structuring included. Once the data is in, the model takes two to three weeks. Vendors who quote shorter timelines usually count from the day your data is already collected and prepared, and often do not offer data collection at all.
Read how CFOs and CEOs use the model, or book a demo and bring one portfolio company's questions.