Framing: KPIs are not ROI
ROI is a ratio:
- Numerator (value): incremental value/profit loyalty creates over a time horizon
- Denominator (cost): total program cost (often dominated by rewards/redemption cost)
KPIs are the hard numbers you can use to operate the program day-to-day and tell an evidence-based story year over year.
The KPI ladder
Use three levels so you don’t confuse symptoms with outcomes:
- Outcome KPIs (prove impact)
- Diagnostic KPIs (explain why outcomes moved)
- Leading indicators (early signals you can act on quickly)
And separate program-level KPIs (durable long term economics) from campaign-level KPIs (what you can test cleanly).
The KPI table
Use this table as a starting point. You don’t need to track everything at once, pick a minimum viable set, define it tightly, and build from there.
| KPI | What It Measures | How to Calculate (Plain Language) | Common Failure Mode | What to Do If It Moves |
|---|---|---|---|---|
| Incremental contribution profit (over X months) | Value created (numerator) | Estimate incremental profit uplift attributable to loyalty over a stated horizon | Using revenue-only; hiding assumptions; claiming certainty | Publish assumptions + sensitivity; pair with tests + KPI evidence |
| Cohort retention (by cycle) | Compounding loyalty effect | % of a cohort that returns and transacts in the next cycle/window | Short windows miss compounding; member vs non-member “proof” | Align windows to repeat cycle; segment; validate against baseline drift |
| Purchase frequency per active member | Repeat intensity | Average/median orders per active member per period | Averages hide mix; promo overlap | Break out by segment; add margin guardrails; watch cannibalization |
| First redemption rate (within X months) | Economic inflection milestone | % of new members who redeem at least once within X months | Counting non-loyalty discounts as “redemption”; gaming with expensive incentives | Fix onboarding + earn/burn pathways; cap incentive cost; monitor cost-per-redeemer |
| Time to first redemption (median / p75) | Speed to value activation | Time from join/first earn to first redemption | Averaging hides tail; ignoring censored members | Use distributions/survival; redesign “path to first reward” |
| Activation rate (first earn within X days) | Early engagement that predicts downstream value | % of enrollees who earn within X days | Treating opens/clicks as activation | Clarify value prop; improve earn accessibility; measure by join channel |
| Ultimate redemption rate (URR) / breakage (by cohort) | True reward cost exposure | Of points issued, what % will redeem (cohort/vintage) | Static averages miss mix shift; optimizing URR as the goal | Track mix shift; forecast URR; manage to value (profit) not URR alone |
| Redemption cost per point / per $ redeemed | Unit cost of benefits | Realized + expected redemption cost per unit, split by reward type | Using industry averages; ignoring catalog mix | Rationalize catalog; renegotiate vendors; steer mix toward efficient rewards |
| Points liability outstanding + expected cost | Denominator risk management | Outstanding issued-not-redeemed value × expected URR × unit cost | Discovering errors late; “spreadsheet” forecasting | Reconcile monthly; scenario-plan promos; update forecasts with behavior changes |
| Points-offer lift (A/B tested) | Causal lift (campaign-level) | Treatment vs control lift in conversion/profit for points-based offers | Claiming it proves full program ROI | Use as a proof point; roll into evidence stack; prevent double counting |
KPI groups
1) Value outcomes (numerator)
Track profit/value outcomes over explicit horizons. If you can’t model incremental profit yet, use directional profit proxies (margin-aware) while you build measurement maturity.
2) Retention + frequency (CLV drivers)
Loyalty value often shows up as compounding retention. These KPIs are critical—but make sure you use the right windows (aligned to your purchase cycle), and avoid member vs non-member stories as proof.
3) Economic inflection points (milestones that change value)
The best operating KPIs are often milestones where customers become meaningfully more valuable (e.g., first redemption). They’re hard numbers you can manage and later map to value impact once your models mature.
4) Redemption economics + liability (denominator realism)
Rewards are often the largest cost component. If you’re not forecasting redemption cost and tracking mix shift, ROI can look “good” until it suddenly doesn’t and the gap accumulates.
A KPI can “improve” while profit gets worse (Example)
You decide to push first redemption aggressively.
- Baseline: 100,000 new members / quarter
- First redemption within 90 days: 10% (10,000 redeemers)
- New incentive: “Redeem any reward and get a bonus” (costly)
- After change: first redemption within 90 days rises to 18% (18,000 redeemers)
But the economics matter:
- Incremental redeemers: 8,000
- Incremental gross profit generated per incremental redeemer (90 days): $12
- Incentive + fulfillment cost per redeemer: $20
Net impact over the window:
- Incremental profit: 8,000 × $12 = $96,000
- Incremental cost: 8,000 × $20 = $160,000
- Net: −$64,000
“Lower redemption rate” can look good… and be a warning sign (Example)
A team celebrates that ultimate redemption rate (URR) dropped:
- Last year URR: 80%
- This year URR: 65%
If you manage URR as the goal, this looks like “cost savings.” But there are two possible stories:
- Good story: you improved reward economics (same engagement, lower unit cost)
- Bad story: engagement fell, fewer members get to meaningful value moments, and long-term value drops
If the URR drop is driven by fewer members ever redeeming (or longer time-to-first redemption), you may be degrading the program’s value engine while the cost metric “improves.”
The minimum viable KPI set
If you can only track 8–12 KPIs, start with a balanced set:
- Value outcomes: incremental contribution profit (or best available profit proxy), cohort retention
- Behavior drivers: purchase frequency, activation rate
- Inflection milestone: first redemption rate within X months, time to first redemption
- Denominator realism: URR/breakage by cohort, redemption cost per point, points liability + expected cost
- Testability: points-offer A/B lift (where feasible)
What not to track as “proof”
These are fine as diagnostics, but don’t lead with them as proof of impact:
- Raw enrollments / sign-ups
- Member vs non-member spend gaps (self-selection)
- Email opens / clicks
- App installs / app opens
- Points issued (cost driver, not value)
- Redemption volume without profit context
If you’re looking to go deeper on how to structure your KPIs, build an evidence-based measurement framework, and stay ahead of evolving loyalty trends, we’ve just released our latest loyalty trends report. It’s a practical guide to help you benchmark your approach, refine your metrics, and set KPIs that reflect real business impact.
