# Customer retention rate: calculate who stayed, not just who joined

Calculate customer retention with a defined opening cohort. Use a worked example, CRM checklist and decision worksheet to separate retention from growth.

Canonical page: https://cpcinsider.com/insights/customer-retention-rate/
Author: Vladlens Kecko
Published: 2026-11-03
Updated: 2026-11-03


**Customer retention rate measures how many existing customers remain customers over a defined period.** It answers a different question from “How many customers did we add?” A growing customer base can still hide losses among the customers you already had.

The [Salesforce customer-retention guide](https://www.salesforce.com/service/digital-customer-engagement-platform/customer-retention/) gives the familiar formula: subtract new customers from the ending customer count, divide by the starting count and multiply by 100. For a reliable calculation, start with customer identities and consistent eligibility rules. The worksheet below is our practical implementation, not a reported client result.

## 1. Define what counts as a retained customer

Choose one unit: an individual buyer, a paying company account or a subscription. Do not mix contacts with companies. Then define the observation period and an “active” rule appropriate to the business. An active subscription and a buyer who orders once a year require different rules.

For example, a contracted maintenance provider could count accounts with a live service agreement on each boundary date. An online retailer might instead measure whether a starting customer cohort orders again within a fixed window. State that repeat-purchase definition explicitly; it is not the same as a subscription snapshot.

Record the rule before inspecting results. Changing inactivity thresholds halfway through the comparison can change the answer without changing customer behaviour.

## 2. Build the opening cohort

A cohort is a group with a shared starting condition. Export eligible customer IDs at the start of the period. Match those same IDs against the eligible customers at the end. Deduplicate merged records and document migrations or reactivations.

The direct calculation is **retained opening customers ÷ opening customers × 100**. New customers do not enter that numerator. If the opening count is zero, the rate is undefined; display “not applicable”, not 0%.

Use a restricted internal worksheet. A public report needs aggregate counts, not names, email addresses or private customer notes. Connect the result to your [CRM measurement workflow](https://cpcinsider.com/insights/crm-conversion-tracking/) without adding personal data to analytics events.

## 3. Check the formula with a fictional example

This example describes a made-up maintenance business over one quarter. All 20 newly acquired accounts are still active at quarter end. There are no merges or reactivations.

| Item | Accounts | Meaning |
| --- | --- | --- |
| Opening active accounts | 100 | The fixed starting cohort |
| New accounts still active at the end | 20 | Excluded from retention |
| Ending active accounts | 110 | Existing plus new |
| Opening accounts still active | 90 | 110 minus 20 |
| Opening accounts lost | 10 | 100 minus 90 |

Retention is **(110 − 20) ÷ 100 × 100 = 90%**. The ending base is 10% larger, while 10 of the original accounts have left. Dividing 110 by 100 gives 110%; that ratio describes the size of the ending base relative to the opening base, not customer retention.

The subtraction shortcut needs care. If customers join and leave within the same period, subtracting every acquisition from the ending base can misstate retained customers. Reconcile the new customers actually included in the ending base, or use the direct opening-cohort ID match. Keep acquisition and churn events separately for reconciliation.

## 4. Complete one decision worksheet

| Field | What to write |
| --- | --- |
| Population and dates | Account type, start date, end date, active rule |
| Opening cohort | Eligible unique IDs and count |
| Retained cohort | Opening IDs that still satisfy the same rule |
| Exceptions | Merges, reactivations, missing records and their treatment |
| Investigation | One observed cancellation pattern; assumptions labelled |
| Action and review | Owner, proposed change, review date and comparison |

Start with a practical question: are customers leaving before onboarding is complete, after a service problem, or at renewal? Use documented reasons where available. “Price” in a cancellation field is a starting point for investigation, not proof that a discount will fix the problem.

## 5. Use the result without confusing the metrics

Customer retention counts customers. Revenue retention concerns revenue from an existing customer group; upgrades and downgrades can move it differently. Neither a website return visit nor an ad conversion establishes that a paying customer was retained.

Add the cohort count, rate and rule to your [marketing scorecard](https://cpcinsider.com/insights/weekly-marketing-scorecard/). For a small cohort, show the counts prominently: losing one of ten customers changes the rate by ten percentage points. Compare equal periods with the same definition and account for seasonal purchasing patterns. There is no universal “good” target in this worksheet.

Before raising acquisition spend, read retention alongside [customer acquisition cost](https://cpcinsider.com/insights/cpc-cpa-cac-differences/) and contribution. A test might improve the onboarding handoff for the next comparable cohort, with an assigned owner and a predefined review window. Treat any difference as an observation requiring context, not automatic causal proof.

## Put the worksheet into a growth decision

The immediate output is one reconciled customer cohort and one question the team can investigate. [Business growth support](https://cpcinsider.com/services/business-growth/) can help connect that question to acquisition, delivery and measurement. [Request a free 30-minute founder-led growth review](https://cpcinsider.com/#audit) for a short written plan with three priorities, with no obligation to hire CPCInsider.