To measure customer loyalty, combine what customers do with what they say. Retention and repeat purchases describe behavior. Recommendation and relationship surveys describe sentiment. Revenue measures help assess commercial value, but none of these alone proves that a customer is loyal.
Start with a small set of measures that fit your business, define their populations and time windows, and use the results to investigate specific customer experiences. This guide explains five core metrics, feedback collection, behavior analysis, and a practical reporting framework. All numerical examples are illustrative.
Define the relationship you want to measure
For a subscription business, continuing to pay and achieving an important product outcome may be useful behavioral signals. For a retailer, repeat purchasing over a suitable buying cycle may be more relevant. An annual purchase should not be evaluated against a weekly-use product’s activity threshold.
Distinguish customers, accounts, subscribers, and individual users. In a business account, the buyer, administrator, and everyday users may have different experiences. A satisfied user is not necessarily the person deciding whether to renew.
Document what counts as an active customer, a qualifying purchase, a cancellation, and a completed survey response. Keep those definitions stable when comparing periods.
| Question | Measure to consider | What it cannot establish alone |
|---|---|---|
| Are existing customers staying? | Cohort retention and churn | Whether they stay from preference, need, or contractual constraints |
| Do customers buy again? | Cohort repeat purchase rate | Why they returned or whether a discount drove the purchase |
| Would customers recommend us? | Net Promoter Score and comments | Actual referrals, future purchasing, or everyone’s opinion |
| What is the relationship worth? | Observed value and estimated lifetime value | Profitability or loyalty without further context |
| Where does friction occur? | Task outcomes, support feedback, and effort measures | The full relationship from one interaction |
1. Calculate customer retention for a defined cohort
A cohort is a group selected using a common starting condition. For a simple customer-retention measure, select customers active at the start of the period and check how many are still active at the end.
Cohort customer retention rate = Starting-cohort customers active at period end ÷ Customers in the starting cohort × 100
For example, if 100 customers are active at the start and 90 of those same customers remain active at the end, retention is 90%. Customers acquired during the period do not enter this starting cohort.
Specify whether you measure end-of-period activity or uninterrupted retention. A customer who leaves and returns can produce different results under those rules. Keep reactivations visible instead of assuming they were continuously retained.
For subscriptions, compare cohorts at the same age, such as their first or third completed month. Stripe’s billing analytics documentation describes subscriber and revenue cohort retention separately, with revenue retention allowing expansion. Check the definitions in your own system before comparing a dashboard with a spreadsheet.
2. Track churn as a distinct view of customer loss
Using the same starting cohort, define customer churn as customers who meet your loss rule during the period divided by the starting cohort. Specify when a loss is recorded: a cancellation request, service end, or another documented event.
In a simple example with no reactivations or ambiguous statuses, losing 10 of 100 starting customers gives 10% churn and 90% retention. That complement relationship requires the same population, period, and mutually exclusive outcome rules.
Separate voluntary cancellations from payment-related losses and other causes where the records support it. A billing failure, a missing capability, and a completed short-term project call for different investigations.
Do not infer dissatisfaction from every departure, or satisfaction from every active subscription. Combine the result with feedback and task outcomes. Our churn diagnosis guide explains how to investigate causes and risk signals.
3. Measure repeat purchases over a suitable follow-up window
A repeat purchase metric needs a clear starting purchase and enough time for another purchase to be plausible. One useful approach is a purchasing cohort: select customers whose qualifying purchase occurred in a specified period, then observe whether they purchase again within a defined follow-up window.
Cohort repeat purchase rate = Customers in the purchasing cohort making another qualifying purchase within the follow-up window ÷ Customers in that purchasing cohort × 100
If 60 of 200 cohort customers make another eligible purchase during the completed window, the rate is 30%. Define how refunds, cancellations, subscriptions, and duplicate orders are treated. Count customers, not the number of orders they placed.
This is a specific cohort definition. Other reports may count customers with multiple purchases inside one reporting period, which answers a different question. Label the method and avoid comparing rates with different buying windows.
Inspect the time to repeat purchase and the role of discounts. Repeat buying can signal a useful relationship, but the measure does not show the customer’s motivation or margin by itself.
4. Calculate NPS from valid survey responses
Net Promoter Score, or NPS, measures stated likelihood of recommendation on a zero-to-ten scale. Bain’s NPS methodology classifies ratings of 9–10 as promoters, 7–8 as passives, and 0–6 as detractors.
NPS = Percentage of valid respondents who are promoters
− Percentage of valid respondents who are detractors
Passives remain in the denominator when calculating both percentages. If 100 valid respondents include 60 promoters, 20 passives, and 20 detractors, NPS is 40. Report it as a score, not “40% loyal customers.” The possible range is −100 to 100.
NPS describes respondents’ recommendation ratings. It is not a direct satisfaction percentage or evidence that they made referrals. Ask a neutral follow-up question about the reason for the rating, and compare comments with the relevant customer experience.
Report the eligible survey audience, invitations, valid responses, response rate, timing, and relevant segments. A change in who responds can affect the score even if the underlying experience is similar. Keep wording, scale, and sampling consistent when reviewing a trend.
5. Distinguish observed customer value from estimated lifetime value
Observed customer value summarizes what has happened so far under a defined revenue or contribution measure. Customer lifetime value estimates future value as well. State whether you mean revenue, gross profit, or another contribution measure.
Stripe’s documentation describes a subscriber LTV estimate using average revenue per user divided by subscriber churn rate. Treat this as a model-based estimate rather than a known outcome for each customer.
For an illustrative monthly revenue model, average revenue of $500 per customer divided by monthly churn of 0.05 gives estimated revenue LTV of $10,000. This simplified estimate assumes a sufficiently stable revenue and churn pattern; it does not include acquisition cost, service cost, margin, or discounting.
A changing price, expansion, or short observation history can make a simple estimate misleading. A zero observed churn rate does not justify an infinite value forecast. Use cohort evidence and explicit assumptions when the business needs a more detailed model.
A high value estimate can help prioritize an investigation, but it does not prove enthusiasm or explain why customers remain.
Build feedback that explains the numbers
Use relationship feedback to understand the broader customer experience and interaction feedback to understand a particular task. Keep their purposes distinct. A support survey after a resolved ticket should not be treated as a complete view of account loyalty.
Qualtrics’ relationship-survey guidance describes collecting metrics over time alongside open feedback. Explain how feedback will be used and offer a follow-up route when your team can act on it.
Ask a small number of focused questions. For example, combine the relevant rating with “What most influenced your answer?” Avoid leading questions that assume customers had a positive experience.
- Sample across relevant customer groups, rather than only your most active users.
- Use a cadence suited to the relationship and avoid repetitive invitations.
- Keep incomplete and missing responses visible.
- Group comments by specific themes and inspect the original wording.
- Assign actions to the team that can address the finding.
Use our feedback analysis guide for organizing qualitative evidence. Automated labels can assist review, but check a sample for misinterpretation.
Use behavior analytics as evidence, not a loyalty label
Choose behaviors tied to customer goals, such as a completed workflow or a repeat purchase. Logins, email opens, and content views can provide context, but their meaning varies by product and role.
A customer using an automated integration may log in rarely while receiving value. A frequent visitor may be struggling with an error. Investigate a change in activity before assuming either loyalty or risk.
If you use a composite health score, document its inputs, weights, missing-data treatment, and purpose. Call it a prioritization tool unless you have validated a stronger interpretation. Compare flagged cases with later outcomes and review false alarms.
Use relevant account context with appropriate access, and avoid collecting unrelated personal details solely to make a profile more extensive.
Create a measurement framework that leads to action
Begin with one behavioral measure, one sentiment measure, and the supporting evidence needed to interpret them. Expand the dashboard only when another measure answers a decision you actually face.
Loyalty measurement worksheet Customer or account population: Business relationship and expected buying/usage cycle: Behavioral measure, formula, and time window: Survey measure, sampling, and response coverage: Revenue/value basis and assumptions: Cohorts and segments for comparison: Observed change and supporting comments: Alternative explanations or missing data: Action, owner, and review date:
Compare similar cohorts and review counts alongside percentages. Investigate meaningful changes in retention, purchasing, or feedback. Set targets using your baseline and an addressable problem, rather than assuming a universal industry threshold.
For an initiative such as clearer onboarding, track the eligible audience, actual intervention, customer outcome, and cost. A before-and-after improvement can have other explanations; use a suitable comparison group where feasible. Review whether the benefit lasts beyond an initial discount or campaign.
Our support metrics guide can help connect loyalty findings with service delivery. The useful result is an evidence-backed action, not a single score presented as the whole relationship.
Frequently asked questions
What is the best customer loyalty metric?
Choose according to the relationship. Retention or repeat purchasing measures behavior, while NPS and relationship feedback measure stated sentiment. Use them together with relevant context.
Is retention proof of loyalty?
No. Customers may stay because they prefer the service, need it, face switching costs, or have a continuing contract. Feedback and customer outcomes help explain the behavior.
Can a higher NPS establish that a retention program worked?
Not alone. Check the respondents, timing, comments, actual customer behavior, and a suitable comparison. A survey trend is evidence to investigate, rather than automatic proof of causation.
