Contribution and Media Efficiency: Beyond ROAS, a Media Budget Tool

Download the workbook: an Excel tool, pre-filled with a fictional retailer’s quarter of media. No form, no email, no gate.

Most media reviews rank campaigns by return on ad spend: attributed sales divided by spend. The campaigns at the top get more budget, the ones at the bottom get less, and the business wonders why profit does not follow.

ROAS leaves out four things that decide whether media makes money. The margin on the sales. The returns and the cost of serving the order. The sales that would have happened without the campaign. And the fact that the average pound and the last pound earn very different amounts.

The problem

Why ROAS misleads

It counts revenue, not money

A ROAS of 5 on a 30% margin line and a ROAS of 5 on a 60% margin line are different businesses. Take away returns, the cost of the order and the media itself, and one may lose money.

It credits sales the campaign did not cause

Brand search and cashback affiliates sit at the end of the journey, where customers who had already decided to buy pass through. Attribution hands them the sale. A holdout usually hands most of it back.

It rises when spend falls

The cheapest clicks are bought first. Cut the budget and ROAS goes up, while sales go down. A campaign can beat its ROAS target and miss its sales target in the same month.

From the field. At one business I inherited a paid search agency that reported ROAS well above target. I had not set the target. In the same period the agency missed the sales target and did not spend the full budget. None of that is a contradiction: spend less, buy only the cheapest clicks, and ROAS goes up. The number it was judged on rewarded exactly that.

The method

From attributed sales to contribution

Paste your attributed results by campaign: spend, orders, gross sales and new customers. The workbook then works down five lines.

Step	What it does
Net sales	Attributed gross sales less returns
Sales it caused	Net sales multiplied by the incrementality factor: the share that would not have happened without the campaign. From your own holdout where one exists; from a stated channel default where it does not
Contribution before media	Margin on those sales, less the variable cost of the orders and of handling returns
Contribution after media	Contribution before media less the spend
Return per £1	Contribution before media divided by spend, for the average pound and for the last one

Every campaign is marked Measured or Assumed, so nobody mistakes a default for evidence. A measured result can be above 1: a prospecting campaign that last-click attribution under-credits may have caused more sales than it was given.

The decision

The last pound, not the average one

Returns fall as spend rises. The workbook gives each channel a response exponent, the same diminishing-returns assumption the Budget Builder uses for paid search, social and display. At 0.65, doubling spend raises the sales a campaign causes by about 57%, not 100%. The last pound then earns the average return multiplied by the exponent.

That is the number that sets the budget. A campaign returning £1.20 of contribution for every £1 on average looks healthy. With an exponent of 0.65, its last pound returns 78p, and every pound added from here loses money. The workbook shows, as an estimate, the spend at which the last pound breaks even, and marks each campaign Scale, Hold, Trim or Cut.

From the field. At one business I ran, we cut paid media by 30%. Topline revenue fell by roughly 20%, and the bottom line was stronger without the spend. The pounds that came out had been costing more than they earned.

The sample

A ROAS of 5.7, and £1.03 back for every pound

From the workbook’s fictional retailer, one quarter, £685,000 of media across twelve campaigns:

5.7

ROAS as reported

£1.03

Contribution per £1, on the sales it caused

79%

Of spend where the last pound loses money

The ranking inverts. Brand search reports the highest ROAS, 15, and a holdout shows it caused a fifth of the sales it was credited with: Trim. Cashback affiliates report 9, cause a fifth of their sales, and lose money: Cut. Paid social prospecting reports one of the lowest ROAS figures in the account, 2.4, and its holdout shows it caused 1.8 times the sales last click gave it: Hold. Ranked by ROAS, the budget would move in exactly the wrong direction.

The step-down test

What happens if the budget is cut

The step-down test cuts 10%, 20% and 30% of total spend, starting with the campaigns whose last pound earns least, and compares the result with the same cut spread evenly across every campaign.

Cut in total spend	Business net sales lost	Contribution after media, least efficient first	Same cut, spread evenly
10%	0.5%	+£55,273	+£20,905
20%	1.7%	+£94,761	+£40,073
30%	3.0%	+£117,929	+£57,203

In the sample, every cut adds contribution, and cutting the least efficient spend first adds more than twice as much as cutting evenly. The cut levels can be changed. Where every cut loses contribution, the workbook says so: the budget is then not the problem to solve.

From the field. At one business, marketing was targeted on traffic. It bought cheap display, hit the target, and conversion fell; the display traffic bounced 80% of the time. The way I explained it to the CEO: marketing had brought five coaches of customers to the car park outside the window, and opened the doors on one. Cheap traffic is not efficient traffic.

Evidence

Tests beat attribution

The Tests sheet takes your holdout and on-off results: the sales a campaign was credited with during the test, and the sales it measurably caused. The measured factor replaces the default for that campaign, and the verdict updates.

A large campaign resting on an assumed factor is not moved on the assumption. If it carries more than 5% of total spend, its verdict becomes Test before you cut, trim or scale. In the sample, £445,000 of the £685,000 rests on assumed factors, and the Overview says so.

Each campaign gets What? So what? Now what?, with the money attached: the return on its average and last pound, the verdict, and the spend to move towards. The workbook recommends; the business decides.

Controls

Checks and testing

Ten checks run on the data, including one most media reports never make: whether the channels together claim more sales than the business made. The others cover negative values, unknown channel types, duplicate campaign names, tests that match no campaign, measured factors outside 0 to 2, exponents that would make the last pound earn as much as the first, and the share of spend measured by a test. The workbook has been tested in 19 scenarios, from an empty sheet to a full cut of the budget, and every cell you should not change is locked.

The series

With the other tools

The Ecommerce Budget Builder plans media by channel with the same diminishing returns; this workbook tells you which campaigns deserve the next pound of that plan. The Ecommerce KPI Framework puts contribution after marketing, acquisition cost, first-order payback and the marginal cost of a new customer on the dashboard as decision metrics; this is where they are calculated campaign by campaign. Promotion Economics asks the same question of a discount: what did it cause, and what would have happened anyway?

Limits

What it does not do

The defaults are assumptions, not benchmarks. The incrementality factors and exponents are starting points. Replace them with your own tests.
The response curve is an estimate. Move spend in steps and measure each one; do not jump to the break-even spend in one go.
It judges campaigns one at a time. Campaigns that work together, such as prospecting that fills the audience retargeting converts, should be tested together.
It counts the first order. Customers who buy again are worth more. That belongs in a lifetime value view built on cohorts, not assumed here.

Sources

Where this comes from

The method and the episodes come from running ecommerce P&Ls and digital marketing budgets of up to £12m since 1999. The figures in the workbook belong to a fictional retailer and describe no real business or benchmark.

If your media report leads with ROAS, run one quarter through the workbook. If more than half your spend sits in campaigns whose last pound loses money, that is worth a conversation before next year’s budget is set.

Back to Thinking