Measure customer retention with the right starting group

Choose the question before the metric

Three common questions require different measures:

• Are existing clients staying? Follow the customers active at the start of a period.
• Do new buyers return? Follow a group whose first purchase happened in the same period.
• Are we keeping the revenue from existing clients? Follow revenue attached to a fixed starting group.

Do not label all three “retention” without further explanation. A subscription business, a project consultancy and a gift retailer have different evidence of an ongoing relationship.

For a consultancy, completing a one-off project as agreed is not automatically a lost customer. For a subscription, an active contract may be a reasonable starting definition. For repeat-purchase retail, a second completed order within a specified window may be the more useful question.

Write a one-sentence metric definition and put it beside the report. If the team cannot agree on that sentence, a chart will not resolve the ambiguity.

Freeze the starting group

Suppose you want to know how many existing subscription customers remained active through a quarter. Identify every qualifying customer at the start, save that group and check those same customers at the end.

Use this calculation: retained starting customers divided by starting customers, multiplied by 100.

Illustrative calculation: 80 customers qualify at the beginning. Of those same customers, 70 remain active at the end. Customer retention is 70 divided by 80, or 87.5%.

If 30 new customers joined during the quarter, the business ends with 100 active customers in this simplified example. That growth does not change the 87.5% retention of the original group. Mixing the newcomers into the numerator would answer a different question.

A common shortcut subtracts new customers from the ending total. That can work under a clean set of definitions, but direct customer-level matching is easier to inspect when there are reactivations, mergers, duplicate accounts or status corrections.

Also record the number behind the percentage. Losing two out of ten clients and twenty out of a hundred produces the same rate but presents different practical work.

Compare new buyers at the same age

A cohort is a group defined by a shared starting event, such as first purchase in a particular month. It helps you compare what happens after customers join.

Give every cohort the same amount of follow-up time. If March buyers have had six months to return and August buyers have had one, comparing their total repeat-purchase rates is unfair.

For an illustrative retail report, define “a second completed, non-cancelled order within 90 days of the first order”. Only include customers whose first purchase was at least 90 days before the reporting cut-off. Show newer groups as not yet mature rather than treating their missing future activity as failure.

Exclude test orders and resolve duplicate identities consistently. Decide whether exchanges, replacement shipments and split orders count before calculating. Usually they should not masquerade as fresh buying decisions, but the right definition depends on your data and question.

Platform reports also have their own definitions. For example, Stripe’s subscriber cohort guidance groups subscribers from when they first generate positive monthly recurring revenue and keeps returning subscribers in their original cohort. Check the definition before comparing a platform chart with your own spreadsheet.

Keep customer retention separate from revenue retention

A business can retain most customers but lose revenue when they reduce their scope. It can also lose several smaller customers while revenue rises because one large client expands.

To see the difference, keep a fixed starting customer group and examine its revenue separately. Exclude revenue from entirely new customers when the purpose is to understand existing-customer retention.

Here is a simplified monthly recurring revenue example, with all amounts illustrative and in Australian dollars:

• The starting group generates $20,000 a month.
• Cancellations remove $2,000 and downgrades remove $1,000.
• Expansion within that same group adds $4,000.
• Gross revenue retention, excluding expansion, is $17,000 divided by $20,000: 85%.
• Net revenue retention, including expansion, is $21,000 divided by $20,000: 105%.

The 105% result does not mean everyone stayed. Keep the customer count and lost revenue visible alongside it. For project work or seasonal sales, recurring-revenue formulas may be inappropriate; use a clearly defined comparison suited to that business instead.

Document the awkward cases

The most important measurement work often sits outside the formula. Give these situations an explicit rule:

• Paused customers: whether a pause qualifies as active, inactive or a separate state.
• Reactivations: whether the report measures continuous retention or active status at the end.
• Multiple locations: whether a customer means an organisation, a site or an individual account.
• Failed payment: when a recoverable billing problem becomes a cancellation in the measure.
• Customer identity changes: how account merges, new emails and business-name changes are reconciled.
• Refunds and corrections: when historical figures are restated and how that is noted.

A customer who left and returned during the quarter may count as active at the end but not continuously retained. Both measures can be useful. Name the one you are reporting instead of switching quietly between them.

If there are no starting customers, show the measure as not applicable. A zero denominator does not produce a meaningful retention percentage.

Turn the result into a decision

Look for patterns within sensible groups: product, service type, acquisition period or onboarding approach. Avoid slicing a small base so finely that one account appears to represent a trend.

If customers leave before their first review, inspect onboarding and expectation-setting. If repeat purchases decline in one category, check availability and product fit. If revenue retention rises while customer retention falls, investigate whether expansion is masking broader losses.

A before-and-after improvement does not by itself prove an email campaign caused the change. Pricing, service quality, stock and seasonality may have changed too. Record those factors and use a suitable comparison where available.

The first useful deliverable can be a modest one-page report: definition, period, starting count, retained count, result, exclusions and next investigation. That is more actionable than a large dashboard with an unexplained percentage.