You’re looking at a spreadsheet trying to figure out whether last quarter’s marketing was worth what you paid for it. Revenue is up. Spend is up. Somewhere in there is a number that tells you whether to do more of this or stop, and you can’t find it.
You searched for how to measure marketing ROI, so here it is, and then here’s the part that actually decides whether the answer means anything.
The formula, so we can get it out of the way
Return on marketing investment is a ratio:
(Revenue attributable to marketing − Marketing cost) ÷ Marketing cost
Spend $10,000, generate $40,000 in attributable revenue, and you’re at 300 percent, or 3:1. Most people running the calculation use gross profit instead of revenue, because a 3:1 return on a product with a 20 percent margin is a loss. Swap gross profit in and the number gets honest fast.
There’s a second version worth knowing. Incremental ROI compares the period with the campaign against a baseline of what the business was already doing, so you’re measuring the lift rather than taking credit for every sale that happened while the campaign was live.
That’s the arithmetic. It’s fourth-grade math, and it is not why your marketing ROI number is useless.
The number is useless because four of the values you just plugged in aren’t measurements. They’re decisions. And in most businesses under $50 million, nobody has made them on purpose.
Input one: what you’re calling a lead
Ask your salesperson how many leads marketing sent last month. Then ask whoever runs marketing. You will frequently get two different numbers, and neither person will be lying.
One of them is counting form fills. The other is counting people who took a call. Somewhere in a tool there’s a third number counting anyone who downloaded anything, and that’s the one feeding the dashboard.
This isn’t a reporting glitch. It’s the absence of a definition. Until someone writes down what event counts as a lead and both sides agree to it, every conversion rate in the business is measuring a different thing than the one next to it, and the ROI calculation inherits all of it.
The fix is unglamorous: one written definition, one owner, one place it lives. That’s what a working marketing system is for.
Input two: what a customer is actually worth
The second number people get wrong is the revenue side, and they usually get it wrong downward.
If your ROI math uses the first sale, you’re valuing a customer at the smallest number available. A client who buys once for $3,000 and a client who buys $3,000 a year for six years enter the calculation identically. Every channel that brings in patient, high-retention buyers looks worse than the channel that brings in one-and-done transactions.
Businesses cut good channels over this. Not because the channel failed, but because the measurement was truncated at the first invoice. It’s a common reason the budget feels like waste when it isn’t.
You don’t need a sophisticated lifetime value model. You need average order value, average repeat purchases, and average retained years, pulled from your own books. Rough is fine. Rough and roughly right beats precise and structurally wrong.
Input three: the window
Marketing spend and marketing revenue don’t happen in the same month.
If your sales cycle is four months and your reporting window is thirty days, you are systematically comparing this month’s cost against last quarter’s return, and the mismatch will make your best long-cycle work look like waste. The attribution window in most analytics tools is set to a default somebody picked, and that default is almost never derived from how long your buyers actually take.
Pull your last thirty closed deals. Measure the days between first touch and signed. That’s your window. Set the reporting period to match it, and accept that a quarterly read on a six-month cycle is a preview, not a verdict.
This is also why the compounding matters. Content, search, and reputation pay out on a curve that a monthly report can’t see, which is a separate argument from whether they’re working.
Input four: the baseline
Here’s the input that gets skipped most, and it’s the one that turns the whole exercise into theater.
“We ran the campaign and revenue went up” is not evidence. Revenue also goes up in your busy season. It goes up when a competitor closes. It goes up when a salesperson finally works the list she’s been sitting on. Attributing the whole increase to the campaign requires knowing what the business would have done without it, and almost nobody establishes that number before spending the money.
You don’t need a controlled experiment. You need a written expectation, recorded before the spend: here’s our baseline run rate, here’s what we think this produces on top of it, here’s the window. Then you compare. A prediction written down in advance is worth more than a sophisticated attribution model applied afterward, because the model can be tuned until it agrees with you and the prediction can’t.
Write the number down before you spend. Most of the diagnostic value of measurement comes from that one habit.
The honest scorecard
Some of this genuinely isn’t measurable, and the sophisticated-sounding move is to pretend otherwise.
A referral from someone who read three of your articles two years ago and finally had a reason to call does not carry a UTM parameter. The deal you won because a prospect already recognized your name will attribute itself to whatever he clicked last, usually a branded search, and branded search will look like a hero.
So build the scorecard in three columns. What you can measure directly, with the definitions above locked: cost per qualified lead, close rate by source, gross profit per channel against its window. What you can measure as a proxy: branded search volume, direct traffic, inbound referral count, win rate on deals where the prospect came in already knowing who you are. And what you can’t measure, named explicitly and left unmeasured rather than assigned a fake number.
That third column is the one that takes discipline. An unmeasurable item assigned a zero is a decision to stop funding it, and that decision usually gets made silently, by a dashboard, on behalf of an owner who never agreed to it.
The point of measuring marketing isn’t a number for the board deck. It’s knowing what to stop doing. A scorecard that can’t tell you that, however precise it looks, hasn’t earned the time you spent building it. Simplicity on the other side of complexity: four definitions, one window, one baseline written in advance, and three honest columns.
If the hard part is that nobody in the building owns those definitions, that’s not a reporting problem to solve with better software. That’s what senior marketing judgment is for, and it’s the first thing fractional marketing leadership puts in place, because nothing downstream of it can be trusted until it exists. It’s also the question to settle before you decide how much to spend next year.
Frequently asked questions
What is a good marketing ROI?
The figure most commonly cited is 5:1, with 2:1 treated as the break-even neighborhood, but treat that as a convention rather than a standard. It depends almost entirely on your gross margin and your sales cycle. A 3:1 return on a 70 percent margin service is healthy. The same 3:1 on a 20 percent margin product loses money. Calculate against gross profit and the question mostly answers itself.
What’s the difference between ROI and ROAS?
ROAS, return on ad spend, divides revenue by advertising cost only. ROI accounts for the full cost of producing the return, including labor, tools, agency fees, and production. ROAS is a channel-level tactical read. ROI is a business-level read. A campaign can post a strong ROAS and still lose money once the people and platforms behind it are counted.
How long should I wait before measuring a marketing campaign?
Match the measurement window to your actual sales cycle, which you can calculate from the days between first touch and closed deal on your last thirty wins. Reading a six-month cycle on a thirty-day window will make good work look like failure. Early reads are useful as directional signals, not as verdicts.
Can you measure brand marketing ROI?
Not directly, and the attempts that claim to usually smuggle in an assumption doing the real work. Use proxies instead: branded search volume, direct traffic, inbound referral counts, and win rates on deals where the prospect arrived already knowing your name. Track the proxies over quarters rather than months, and name the item as a proxy rather than presenting it as attributed revenue.