What is marketing measurement?
Marketing measurement is the practice of estimating what marketing actually caused: which spend, channels and creative produced results the business would not otherwise have had, and at what cost.
Reporting describes what happened. Measurement asks how much of it marketing was responsible for. A campaign can report many conversions and still have caused few of them, if most of those customers were going to buy anyway.
- 01QuestionWrite down the decision it informs
- 02MethodChoose the one that answers it
- 03CheckCompare direction across methods
The main marketing measurement methods
No single method answers every question. Each one sees part of the picture and has a known way of misleading.
| Method | Question it answers | Data it needs | When it misleads |
|---|---|---|---|
| Platform-reported attribution | How much credit does this platform give itself? | Conversion events tracked inside one ad platform | When platforms are added together, because several can credit the same sale |
| Multi-touch attribution (MTA) | Which tracked touchpoints led to a conversion? | User-level journeys across channels | When tracking is incomplete across devices, consent choices or offline sales |
| Marketing mix modeling (MMM) | How much did each channel contribute over time, including offline and brand effects? | A long history of spend and outcomes, plus seasonality, pricing and promotions | When spend barely varies, so the model cannot separate channels, or when it is never checked against experiments |
| Incrementality testing | What would have happened without this spend? | A test group and a comparable holdout group or region | When the test is too small or too short, or the holdout is not truly comparable |
| Post-purchase surveys | Where do customers say they heard about us? | A short question at checkout or signup | When memory favors the most recent or most memorable channel |
Different methods should agree on direction. Expecting them to agree on the number is how teams talk themselves out of good evidence.
Effectiveness vs efficiency
Effectiveness asks whether marketing produced the outcome at all. Efficiency asks what each result cost. A channel can look efficient on platform-reported numbers and still be ineffective, if it mostly reaches people who would have bought anyway.
Two formulas connect them: incremental lift = (test conversion rate - holdout conversion rate) / holdout conversion rate, and incremental return on ad spend (iROAS) = incremental revenue / spend.
Illustrative example: a region that sees the ads converts 1,200 people per 100,000, and a comparable holdout region converts 1,000 per 100,000. Lift is 20 percent, or 200 incremental conversions per 100,000 people. If that cost $20,000 and the average order is $100, incremental revenue is $20,000 and iROAS is 1.0, even if the platform reports a return several times higher.
How to choose a marketing measurement method
Start from the decision. A team that writes the question down first usually finds it needs only two methods.
- Write down the decision the measurement informs, such as next quarter's channel budget or whether to keep a campaign running.
- Use platform data for day-to-day optimization inside one channel, where its blind spots matter least.
- Use mix modeling for planning across channels and periods, including channels that do not produce clicks.
- Run an incrementality test before a large budget change depends on one channel's reported results.
- Agree on definitions, attribution windows and reporting periods before the period starts, and record them beside the results.
When measurement methods disagree
They will. Each method counts a different thing, so their totals differ by design. The useful check is direction: if platform data, the mix model and an experiment all say a channel is improving, the evidence is strong even though the sizes differ.
When they point different ways, look for the cause before choosing a favorite. A tracking change, a shifted attribution window or a short test explains most disagreements. Experiments are the strongest tie-breaker, and many teams use them to calibrate their mix model over time.
Where marketing measurement goes wrong
Most mistakes come from the inputs rather than the method:
- Adding conversions across platforms, which counts the same customer more than once. Spend can be summed across channels; platform conversions and revenue usually cannot.
- Changing a conversion definition or attribution window in the middle of a period and comparing across the change.
- Reading a test before it has run long enough to separate a real difference from noise.
- Judging upper-funnel channels only on short-window, last-click results, which they rarely win.
Frequently asked questions
What is the difference between marketing analytics and marketing measurement?
Analytics describes and explores what happened in the data. Measurement estimates how much of the outcome marketing caused. Good measurement depends on good analytics, but answers a narrower question.
Is marketing mix modeling better than multi-touch attribution?
Neither is better in general. Mix modeling suits budget planning across channels, including offline ones. Multi-touch attribution suits faster optimization where journeys are well tracked. Many teams use both and check them against experiments.
How long should an incrementality test run?
Long enough to cover at least one full purchase cycle and to collect enough conversions that the difference between groups is larger than normal week-to-week variation. Decide the length before the test starts, not when the result looks good.
What is measurement triangulation?
Using several methods together and comparing them, rather than trusting one. Agreement on direction across methods is the signal to act on.
