The Marketing ROI Formula, and the Revenue That Was Coming Anyway

The agency deck said the spring campaign returned 257 percent. The finance lead, working from the same campaign, said it lost money. Neither of them made an arithmetic error. Both used the marketing ROI formula, and both got a number that follows correctly from the inputs they chose. The argument was never about the formula. It was about which revenue is allowed into it.

That is the whole problem with this metric, and it explains why marketing ROI is the number most often quoted and least often believed. The formula has one subtraction and one division. The subtraction is where the dispute lives, because it asks for the profit the campaign produced, and a campaign report can only show you the profit the campaign was credited with. Those are different figures. On most accounts the gap between them is wide enough to flip the sign.

The formula, and three campaigns hiding in one report

Written out properly, the marketing ROI calculation compares the profit a campaign added with everything it cost to add it. The shortcut version sits underneath it for comparison.

Marketing ROI = (Incremental gross profit − Marketing cost) / Marketing cost
Shortcut = (Revenue − Marketing cost) / Marketing cost

Marketing cost = $30,000 media + $6,120 agency, creative and tools = $36,120
Profit per order = $86.00 order x 42% contribution = $36.12

The version most pages quote puts revenue where incremental gross profit should be. The marketing ROI calculator already swaps revenue for gross profit and explains which costs belong in the denominator, so this post takes that step as given and goes after the word the calculator cannot fill in for you: incremental. Here is one campaign run through all three readings. The ad platform credited it with 1,500 orders. A holdout test, described further down, put the real figure at 900.

Same spend, same month, same store. The first line is what goes in the deck. The second is what a careful marketer reports. The third is what happened to the bank balance, and it is a loss of $3,612.

One detail in the third line is where a lot of campaigns hide. On media alone the incremental orders cleared the bar. $77,400 of incremental revenue on $30,000 of spend is a ROAS of 2.58, above the 2.381 break-even ROAS a 42 percent margin requires, the line worked out in the guide to what a good ROAS is. The media made $2,508. The $6,120 of agency fees, creative and tools turned that into a loss. A ROAS test cannot see that, which is why ROI exists as a separate number.

How much credited revenue can be baseline before you lose money

You will rarely know the incremental figure precisely. You do not need it precisely to make a decision, only whether it sits above or below one threshold, and that threshold falls straight out of the formula.

Tolerable baseline share = 1 − 1 / (1 + ROI on credited gross profit)

At 50% credited ROI: 1 − 1 / 1.50 = 33.3%

Read it like this: a campaign that shows 50 percent ROI on credited gross profit stays profitable only if fewer than a third of the orders it was credited with would have arrived without it. Put another way, the campaign needed 1,000 incremental orders to cover $36,120 at $36.12 each. It was credited with 1,500, so it could afford to be wrong about 500 of them. The holdout said it was wrong about 600.

The table turns an attribution argument, which nobody wins, into a yes or no question. Retargeting audiences and branded search are made largely of people who were already on their way to you, so a baseline share above a third should surprise nobody on either line. Prospecting aimed at cold audiences usually carries a much smaller one. The same 50 percent credited ROI can therefore be comfortable on one line of the media plan and a loss on the next, and the ROI report will print them as equals.

This is the problem the marketing efficiency ratio has, in a different costume. MER blends organic revenue into its numerator and falls with scale for reasons that have nothing to do with the ads. Credited ROI blends baseline revenue into its numerator and looks better than the campaign deserves for the same reason. The MER guide works through that version.

Measuring the baseline without an attribution argument

The honest way to get the incremental figure is to withhold the campaign from part of the market and count what you lose. On this account the store matched two sets of regions on the previous quarter's orders, ran the campaign in one set for four weeks and kept it dark in the other. The test regions took 4,100 orders. The control regions took 3,200. The difference, 900, is what the campaign did. The platform, which counts every order it touched, reported 1,500 for the same regions over the same weeks.

Two cautions keep a holdout from producing a confident wrong answer. The first is size. A gap of 900 on a base of 3,200 is easy to read, but a gap of 90 would sit inside ordinary week-to-week noise, so check the split with the A/B test calculator before you believe it. The second is spillover: people in control regions see national TV, hear about you from friends in test regions and search for you anyway. That leaks lift into the control and makes the campaign look smaller than it is. The error runs in the conservative direction, which is the right one for a budget decision.

The same correction shows up in acquisition cost. $36,120 over 1,500 credited orders is a CPA of $24.08, comfortably under the $36.12 each order is worth. Over 900 incremental orders it is $40.13, above it. Feed the CPA calculator both order counts and it will show you both verdicts.

There is a cheap version of all this that plenty of businesses use without realising it is a holdout at all. A formula that turns up on many of the pages ranking for this query subtracts marketing cost from sales growth, which quietly treats last month as the control group. That works when last month is a fair stand-in for this one. Run it across a seasonal peak and it will credit your campaign with the calendar.

The window decides the sign as often as the baseline does

The -10 percent above is measured at the first order. It assumes a new customer buys once and never returns, which is false for nearly every business running prospecting, and it is where the finance lead in the opening overstated the case in the other direction. Assume all 900 incremental orders were first purchases from new customers, and that reorders follow the store's history: 30 percent buy again within six months and 45 percent within twelve. Count one repeat per returning customer, at the same $86.00 order and 42 percent contribution.

The campaign crosses break-even once 100 of the 900 have come back, roughly one customer in nine. For reporting, that means a marketing ROI without a window is a range rather than a number, and two campaigns measured over different windows cannot be compared at all. Label the figure with its window every time: incremental, six-month, 17 percent. The LTV calculator handles the longer view, including the discounting this table leaves out for clarity, and the guide to CAC payback covers how long you can afford to wait.

One boundary. Only the repeat orders of incremental customers count. The 600 baseline customers would have reordered too, and folding their repeat purchases in is the credited-revenue mistake made again at a longer horizon.

What to do with this

Take the largest campaign from last quarter and compute its ROI on credited gross profit. Convert that into the baseline share it can survive with 1 − 1 / (1 + ROI). Then ask whoever owns the campaign what evidence says the real baseline is below that share. If the answer is a platform report, book a four-week regional holdout on that campaign before the next budget round, and set its budget from the incremental figure the test returns, measured over a window you wrote down before the test started.

Figures here come from one worked example built so that every intermediate reproduces the printed result. They are illustrative rather than a survey, and the baseline share on your account is something only your own holdout can tell you.

Related reading: why MER falls as spend rises and profit on ad spend, and when it beats ROAS.