Marketing return on investment: the formula, the version finance accepts, and how to prove it

Marketing return on investment is gross profit generated by the marketing, minus the cost of the marketing, divided by the cost of the marketing, expressed as a percentage or a multiple. The number that survives a conversation with finance uses gross profit rather than revenue, includes every cost of running the activity rather than just the media spend, and states the time window it covers. A return calculated on revenue will be several times too flattering, which is the single most common reason a marketing ROI figure gets rejected.

Worked through: spend 20,000 on a campaign, attribute 100,000 of revenue to it, and at a 40 per cent gross margin that revenue carries 40,000 of gross profit. The return is 40,000 minus 20,000, divided by 20,000, which is 100 per cent, or a multiple of two. The same campaign calculated on revenue would read as 400 per cent, and that is the number nobody in finance will sign.

What goes in the two halves

In the return	In the investment
Gross profit from orders attributed to the activity	Media and platform spend
Incremental repeat purchases inside the window, if you can trace them	Agency fees and retainers
Contracted recurring revenue, at its gross margin, over the stated term	Production: creative, photography, video, print
	Software and data used for the activity
	The salary share of the people who ran it
	Discounts and offer costs given to win the sale

Two decisions to make once and then keep, because changing them between reports is what makes a trend meaningless. First, whether salaries are in or out. In is more honest and lower; out is more common and makes channels comparable. Pick one, write it at the top of the report, and keep it for the year. Second, your attribution window, which should reflect the real length of your sales cycle rather than a platform default.

ROI, ROAS and MER, and when each is the right one

• ROI, or ROMI. Profit based, and the right measure for a board, a budget request or a decision about whether an activity should continue.
• ROAS. Revenue divided by ad spend, read as a multiple. Useful inside an ad account for comparing campaigns day to day, because it is fast and the platform calculates it. It ignores margin and every cost that is not media, so it never travels to a board paper on its own.
• MER, or blended ROAS. Total revenue divided by total marketing spend. This is the number that catches what attribution misses, which makes it the useful sanity check on top of everything else.
• CAC and payback period. The right measures for subscription and repeat purchase businesses, where a single transaction ROI understates the value. Payback in months is often a more useful planning number than a percentage.
• Contribution margin after marketing. Gross profit minus marketing cost in absolute money. Finance teams tend to prefer this because it is a figure in the accounts rather than a ratio.

Measuring it by channel, honestly

Different channels can be measured with different confidence, and pretending otherwise is where marketing ROI reporting loses credibility. A workable approach by channel type:

• Paid search and paid social. Platform data for in flight decisions, then a holdout or geographic split test each quarter to establish what the platform numbers are worth. Platform reported conversions are the channel's own account of itself; a test is the independent one.
• Email and owned channels. Measurable with a control group. Hold back a random slice of the list and the difference between the two groups is the return.
• Organic search and content. Measure on a cohort. Take the pages published in a quarter and track the gross profit from their sessions over the following twelve months, against the full cost of producing them.
• Brand and offline. Measure on market level evidence rather than clicks: branded search volume, direct traffic, enquiry volume by region, and a geographic test where you run the activity in some areas and not others.
• Events and field marketing. Pipeline created and closed within a named window, with the cost of attendance and staff time included.

The test that turns a correlation into a return

Attribution tells you which touchpoints were present before a sale. A test tells you which spend caused it, and the difference between the two is often large. Three practical designs:

1. Geographic holdout. Run the activity in a matched set of regions and pause it in another, for long enough to clear your sales cycle. Compare profit per region, not clicks.
2. Audience holdout. Exclude a random percentage of the target audience and compare conversion rates between held out and exposed.
3. Scheduled on and off. Alternate periods with and without the activity, repeated enough times that seasonality cancels out.

Run one test per quarter on your largest line of spend. The finding is usually worth more than the entire rest of the reporting, because it tells you the one thing a dashboard cannot: what would have happened anyway.

This page covers the calculation and the proof. For attribution specifically, see attribution explained and a measurement plan you can keep; for the cost side see customer acquisition cost; and for presenting the spend see the marketing budget format finance will approve.

Asked and answered: marketing return on investment

What is a good marketing ROI?

It depends on your gross margin and your sales cycle, so the comparison that matters is against your own alternatives: your other channels, your cost of capital, and what the money would earn elsewhere in the business. A return of five to one on gross profit is strong in most consumer categories and unremarkable in high margin software. Set the threshold with finance before the campaign, and the reporting conversation afterwards becomes simple.

Should ROI be calculated on revenue or profit?

Profit, specifically gross profit, which is revenue minus the direct cost of delivering it. Revenue based figures are the main source of inflated marketing claims, and finance will discount any number built that way. Keep a revenue version alongside if you like, clearly labelled as ROAS rather than ROI.

How do I handle a long sales cycle?

Measure in cohorts rather than calendar months. Group the leads created in a month and follow that group until the cycle completes, so the return lands against the period that generated it. Report pipeline created as the leading indicator and closed profit as the lagging one, and state both windows on the report.

What about brand marketing that has no trackable click?

Measure it at market level with a geographic test, and watch branded search volume, direct traffic and enquiry volume as supporting evidence. Brand activity is best judged over quarters rather than weeks, so set the review period before you start rather than after somebody asks.

How often should marketing ROI be reported?

Monthly at the channel level for operational decisions, quarterly at the programme level for budget decisions, and annually with the test results included. Keep the definitions and the window identical each time, because a change of method mid year costs you the trend, which is the most valuable part of the whole exercise.

The return your model files under direct

Every method above traces a visit, a click or a region. A buyer who asked ChatGPT or Perplexity for a shortlist and then typed your name into the address bar leaves none of that. The model books the sale as direct traffic, and the work that got you named never appears as a cost or a return.

You can make it visible. Ask the five questions your buyers ask before choosing, in two or three assistants, with your brand name left out, and record who is named. AI Knows Us does this across the major assistants and keeps the sources alongside.