How to measure marketing ROI when you can't track everything

Aug 18
3 min read

Updated: 3 days ago

Introduction

Marketing return on investment sounds like a solved problem. Divide what you earned by what you spent. In practice almost every business gets it wrong in one of three ways: using revenue instead of profit, attributing sales to whichever channel touched them last, or quietly ignoring the channels that cannot be tracked.

Each error makes marketing look better or worse than it is, and both directions lead to bad spending decisions.

1. The formula, with the correction that matters

Marketing ROI = (gross profit from marketing − marketing cost) ÷ marketing cost

The word doing the work is gross profit, not revenue.

If a campaign costs $2,000 and produces $8,000 of revenue at 30% margin, the profit is $2,400. ROI is (2,400 − 2,000) ÷ 2,000 = 20%. Using revenue instead would have suggested 300%, and a business acting on that figure would scale a campaign that barely covers itself.

2. Include the costs that do not arrive as invoices

Marketing cost is not only media spend. Include tools, fees paid to agencies or freelancers, and a realistic value for staff time.

Include first-purchase discounts too. A promotional price given to acquire someone is a marketing cost even though it appears as reduced revenue rather than as spend.

Businesses that count only ad spend routinely overstate ROI by a wide margin.

3. How to measure marketing ROI when attribution is imperfect

Attribution will never be complete. Someone hears about you from a friend, sees an advert twice, searches your name, and buys. Which channel earned that?

The practical answer is to pick a rule, apply it consistently, and treat the output as comparative rather than absolute. You are not trying to discover objective truth about each sale. You are trying to compare this month to last month on a consistent basis.

A workable default for a small business is first-touch for understanding what creates demand, and last-touch for understanding what closes it. Pick whichever question matters more to you right now, and do not switch mid-year — a change of rule looks exactly like a change in performance.

4. Ask customers, because it is cheap and it works

The single most underused attribution tool is a question at the point of purchase: how did you hear about us?

The answers are imperfect — people misremember — but they capture what analytics cannot see: word of mouth, offline mentions, the recommendation that started everything. Recorded consistently over months, the pattern is genuinely informative.

Keep it optional and quick, and record it somewhere it can be counted.

5. Use holdouts for channels that resist tracking

Some spending cannot be attributed by any tool: sponsorship, print, local visibility. Rather than assume it works, test it by absence.

Pause it in one location or one period and watch what happens to overall demand. If nothing changes, you have your answer. If it drops, you have measured the contribution more honestly than any tracking pixel could.

This is uncomfortable and it is the only reliable method available for untrackable channels.

6. Choose a sensible period

Measuring ROI weekly produces noise. Measuring annually produces regret.

Match the period to your buying cycle. If customers typically decide within days, monthly reporting is fine. If your cycle runs over months — as with education enrolments or considered B2B purchases — a month of data will consistently understate return, because the revenue lands outside the window in which the cost appeared.

7. Judge ROI against lifetime value, not the first sale

A campaign that breaks even on the first purchase but produces customers who buy repeatedly is profitable. One that returns 200% on first purchase from customers who never come back may not be.

Where you can, calculate ROI using the profit a customer produces over their whole relationship rather than the initial transaction. This single change reverses the apparent ranking of channels more often than any other adjustment.

Conclusion

Measure with gross profit rather than revenue, count all the costs including discounts and staff time, pick one attribution rule and hold it steady, ask customers directly, test untrackable channels by pausing them, and judge against lifetime value where you can.

None of this produces a perfect number. It produces a consistent one — and consistency is what lets you tell whether a decision improved things, which is the only reason to measure at all.