Stop counting revenue that would have happened anyway.

Start with the question

Ask what the program changed compared with what would have happened without it. A holdout group gives you a reference point: eligible customers who do not receive that specific treatment during the test.

Build a clean comparison

Define eligibility before anyone receives the message. Randomly assign customers to treatment and holdout, keep the groups comparable, and measure both over the same period. Avoid letting other campaigns target the holdout in a way that erases the difference you are trying to isolate.

• Choose one primary outcome before launch: repeat purchase, retained active customer or contribution margin.
• Track the number of eligible customers in both groups, including those who never open or click.
• Record offer, channel and incentive cost as part of the result.

Read value after cost

A useful starting calculation is: (value per eligible customer in treatment − value per eligible customer in holdout) × the eligible population. Then subtract the cost of rewards, discounts and delivery. For a commercial decision, contribution margin is often more useful than gross revenue.

Small samples can swing widely. Treat the result as a range of plausible effects, not a precise headline number. Repeat the test or extend it when the decision is large and the signal is weak.

Make the next decision

If a program produces measurable incremental value, look for the segments and moments where it works best. If it does not, change the audience, offer or timing before increasing volume. That is how measurement becomes a growth tool instead of a reporting exercise.