Perksmith · measuring incremental reward impact
The offer was used. Did it bring extra business?
Redemptions tell you customers used the offer. Attributed sales connect purchases to a campaign under your reporting rule. To understand incremental impact, ask what the reward changed: which purchases happened because of it and would otherwise not have happened?
You need a credible comparison, a defined measurement window and a clear account of the costs to answer that question.
Choose the result you want to measure.
Choose one primary outcome: repeat-purchase rate, orders per eligible customer or the amount left after relevant costs, for example. Define who is eligible, when the measurement starts and how long you will observe behavior. Where feasible, randomly assign eligible customers to a reward group and a hold-back group that does not receive the campaign reward.
Keep the groups comparable and analyze customers according to their assignment. Measuring only customers who used the reward selects a particularly engaged group and can overstate the effect.
Compare customers over the same window.
Hypothetical example: 1,000 eligible customers receive a reward and 1,000 form a randomly assigned hold-back group. Over the same window, 120 in the reward group purchase again, compared with 100 in the hold-back group.
The observed repeat-purchase rates are 12% and 10%. The difference is 2 percentage points, equivalent to 20% relative lift over the 10% baseline. That is an estimate with uncertainty, not proof that every reward recipient changed behavior. The example is arithmetic, not a Perksmith customer outcome or statistical-significance claim.
Count the costs as well as the purchases.
More revenue does not automatically mean more profit. Account for margin, reward cost, messaging charges, refunds and the operating costs relevant to your decision. Also check whether the offer brought a purchase forward instead of adding one. A longer observation window can help reveal that difference.
Do not describe sales divided by reward cost as a proven return on investment caused by the campaign. The revenue figure and comparison must support that claim. Make attribution rules, estimates and measured comparisons visible in the report.
Decide whether the evidence is strong enough.
Review group assignment, sample size, customers receiving an offer intended for the other group, missing records, campaigns running at the same time and seasonal changes. If the evidence is inconclusive, say so. If a randomized hold-back is unsuitable, document the alternative comparison and its limitations; do not give it the certainty of a controlled experiment.
Common questions.
Can redemptions still be useful?
Yes. They help explain participation and cost. They do not establish what would have happened without the reward.
Does a positive comparison guarantee profit?
No. Review uncertainty, the amount left after costs and the costs included before making that decision.
What should a pilot agree in advance?
The eligible customers, main result to measure, comparison group, observation period, costs to include and conditions for continuing, changing or stopping.
Everyday sales. Lasting relationships.