Loyalty Program KPIs: How to Choose and Use the Right Metrics

Loyalty program KPIs are the small set of measures a program team uses to judge whether the program is doing its job. They cover whether members join and stay active, what the program costs, and whether it changes behavior enough to pay for itself. Each KPI should be tied to the program's objective, owned by a named person, and reviewed on a fixed schedule.

Many programs collect plenty of data. The harder part is turning it into a few measures that lead to decisions. McKinsey's 2021 loyalty research describes the problem: "Unclear key performance indicators (KPIs), complicated ROI calculations, and the need to account for the balance-sheet impact of liabilities all make tracking toward a healthy and sustainable program complex." A program that reports enrollments and points issued but cannot say what it changed is hard to defend when budgets tighten.

This guide covers how to choose KPIs and run them. It goes from the objective, through the five dimensions of program health, to leading and lagging pairs, vanity metrics, fair comparisons and the review rhythm. It does not repeat metric formulas; those are in the loyalty KPI dashboard post. Proving the program's financial return is covered in the guide to the value and ROI of loyalty programs. One principle runs throughout: measure against the objective, not everything that can be counted.

Choose KPIs from the top down. Begin with what the program is meant to achieve, then pick the measures that show progress toward it. Starting from the data you happen to have tends to produce a dashboard that describes activity without guiding any decision.

A simple way to do this is a KPI tree with four levels:

The guardrails matter because a KPI that becomes a target can invite shortcuts. If store staff are paid on enrollments, sign-ups can rise while many of the new accounts are never used again. Pairing each target with a guardrail, such as the share of new members active after 30 days, makes that kind of gaming easier to see, though guardrails need checking too.

As an illustration: a specialty retailer's objective is more second purchases from first-time buyers. Its outcome KPI is the second-purchase rate within 90 days, compared with a holdout that does not get the program's welcome offer. Its drivers are the share of new members who click through the welcome message and the share who reach a first reward. Its guardrails are offer cost per new member and margin on second orders. A handful of numbers like these, tied to one objective, tell the team more than a page of totals.

A program's health has more than one side, and a team that watches only spend sees results without causes. Five dimensions cover most programs.

Each dimension needs at least one KPI, but not every KPI deserves a place on the dashboard. Pick one or two per dimension that connect to the objective, and keep the rest as diagnostics to check when a headline number moves.

A few cautions apply to specific dimensions:

• Economic: the gap between member and non-member spend is a description, not a result. Customers who choose to join often already buy more, so the gap overstates what the program causes. The same applies inside the program: McKinsey's 2021 research reports that active members spend 10% more than enrolled but inactive members, and redeemers 25% more, a correlation that does not show the program caused the difference. A program can also raise member activity and still lose money when reward cost per member grows faster than the incremental margin it produces, which is why reward cost sits beside incremental margin in the table. The economic KPI that matters is incremental margin against a comparison group (see Set Baselines, Targets and Comparisons below). Outstanding points liability belongs here too; agree with finance how it is estimated and reported. Member lifetime value carries the same caution: compare it with a holdout or matched group before crediting the program.
• Engagement: breakage, the share of points never redeemed, reads two ways. A rising rate lowers reward cost, but it can also mean members are losing interest. Track it alongside the active member rate, and treat it as a diagnostic rather than a target.
• Emotional: Net Promoter Score and survey sentiment are signals, not proof of loyalty. A single survey question is sensitive to who answers and when. Before treating either as a KPI, check whether members with higher scores actually stay longer or buy more in your own data.

B2B programs need two layers: account-level KPIs, such as share of the account's purchases and number of product lines bought, and user-level KPIs, such as training completions and portal activity. A channel program with a few hundred accounts will also see larger swings from period to period, so read its trends over longer windows.

Lagging indicators, such as revenue, retention and incremental margin, confirm what has already happened. They are the numbers leadership cares about, but they move slowly and arrive too late to fix the period they describe. Leading indicators move earlier and give the team time to act. A useful dashboard pairs each lagging KPI with one or two leading ones.

A leading indicator is only as good as its link to the outcome, and that link varies by program. Check each pairing against your own past cohorts before you rely on it, and drop any that do not hold. A new program has no past cohorts, so treat its pairings as hypotheses and confirm or drop them as the first cohorts mature.

Forrester's data suggests members can disengage without leaving, though the figures predate 2022. Its June 2022 data snapshot reports that in 2021, 38% of US online adults agreed they frequently forget to use the loyalty programs they belong to, an increase of 15 percentage points from 2019. A program that tracks only lagging numbers may not notice members drifting away until retention has already fallen.

A vanity metric looks good and changes no decision. Total enrollments, points issued and app downloads all tend to rise over time whether or not the program is working. A simple test catches many of them: ask what it would take for the number to rise while the business got no benefit. If the answer is easy, the number should not be reported as success on its own; pair it with a guardrail or a comparison group, or move it to the diagnostics.

On open rates specifically, Apple's Mail Privacy Protection "prevents senders from seeing if you've opened the email message they sent you" (Apple), so open rates no longer show reliably whether members read a message.

Vanity metrics still have uses. Enrollment volume can show whether a sign-up flow is broken, and downloads can show whether an app launch reached anyone. The problem comes when they are reported as success. The guide to why loyalty programs fail covers how programs that look healthy on these numbers can still lose money.

The KPIs a team chooses decide whether it can manage its loyalty program or only report on it. Start from the objective and build down to a few outcome, driver and guardrail measures. Cover all five dimensions, pair each lagging KPI with leading ones you have checked in your own data, and replace vanity metrics with measures that are hard to move unless the program delivers. Set baselines and fair comparisons before anything changes, give every KPI an owner, and review them on a fixed rhythm. A program measured this way can show, or at least estimate, what it changes, and can be steered when it stops.

The case, in numbers