AI Retention Reports: Read Gross and Net Together

The AI-generated update celebrates retention above 100%. On the same page, the customer team reports a cancellation and a downgrade. Those statements can both be correct. The retention figure may include expansion from customers who increased their recurring spend, offsetting losses elsewhere.

For a subscription business, the useful question is which changes a retention metric includes. Gross retention and net retention describe different views of the original revenue base. Reading them together makes a business update more informative than asking one percentage to explain every customer movement.

Start with the metric’s definition

ChartMogul’s gross MRR retention documentation, updated on 22 October 2026, starts with recurring revenue from existing subscribers and subtracts contraction and churn. It excludes expansion, reactivation and new business. Its standard calculation divides the remaining amount by starting MRR.

Its net MRR retention documentation, updated the same day, adds expansion and reactivation while subtracting contraction and churn, again dividing by starting MRR. New business remains excluded. These documented definitions establish why the two figures can differ; they do not establish the performance of any particular business.

MRR means monthly recurring revenue in this discussion. The worked example treats it as a stated recurring monthly amount at two snapshots. It is not an account of cash received during the month, and it includes no one-off charges.

A fictional ledger shows both sides

Imagine a small subscription service with five existing customers at the beginning of a completed reporting month. The starting MRR totals £500. Each customer makes at most one change in this deliberately simplified example.

• Customer A starts at £75 MRR and cancels, ending at £0.
• Customer B starts at £125 and downgrades to £100.
• Customer C starts at £100 and upgrades to £250.
• Customers D and E each remain at £100.

The original customer group ends at £550 MRR: £0 + £100 + £250 + £100 + £100. Churn removed £75, contraction removed £25, and expansion added £150. There are no reactivations, exchange-rate changes, refunds or other adjustments in the example.

Gross retention excludes the £150 expansion. Its calculation is (£500 − £75 − £25) ÷ £500 × 100, giving 80%. It describes the original recurring revenue remaining after the specified losses, before counting expansion.

Net retention includes the expansion. Its calculation is (£500 − £75 − £25 + £150) ÷ £500 × 100, giving 110%. The arithmetic and customer totals were checked for this article. No subscription platform or AI reporting product was tested.

The two figures describe the same invented ledger. The 110% result does not mean every customer stayed, and it does not remove the £100 of recurring revenue lost through cancellation and downgrade. The 80% result does not mean the business had no successful expansion.

Keep new customers outside the retention calculation

Suppose a new customer joins during the month with £200 MRR. The service’s overall ending MRR becomes £750. The original group still ends at £550. Under the stated definitions, that new £200 belongs in the explanation of overall growth, not the numerator of either retention measure.

Dividing £750 by the original £500 would produce 150%, but it would combine the original group with new business. Calling that number net retention would answer a different question from the one the report claims to measure.

For this example, a clear update could state that total MRR rose from £500 to £750, comprising a £50 net increase from the starting group and £200 from new business. It could then give gross retention of 80% and net retention of 110%, with the cancellation, downgrade and upgrade shown separately.

A percentage does not explain a customer’s reason

The fictional ledger identifies what changed in recurring amounts. It says nothing about why Customer A left, whether Customer B needed fewer services or what prompted Customer C’s upgrade. Keep those questions open until there is evidence addressing them.

An assistant should not transform the upgrade into a claim that all customers are more satisfied. Nor should it describe the downgrade as proof of a product failure. In this proposed review, the owner would seek relevant customer feedback and operational records before choosing an explanation.

The next action should reflect the question. A cancellation-reason review, a check of a downgraded plan’s suitability and an investigation of an upgraded account’s needs are different pieces of work. A single retention headline cannot decide which response is appropriate.

Ask for the ledger behind the narrative

An assistant could help draft a report from a checked customer ledger and explicit metric definitions. Ask it to list what belongs to the starting group, separate each movement and show the arithmetic. This is an editorial workflow proposal, not a claim about a particular product’s ability to reconcile your subscriptions.

• State the reporting period and the starting customer group.
• Show starting and ending recurring amounts for that same group.
• Separate cancellations, downgrades, upgrades, reactivations and new business.
• Name the gross and net formulas actually used.
• Keep observations about amounts separate from explanations about customer behaviour.

For the fictional service, the reviewer can trace Customer C’s £150 expansion directly to the difference between its £100 starting amount and £250 ending amount. The same check links A’s £75 loss and B’s £25 reduction to the ledger. That is more reviewable than a narrative containing only the final percentages.

Read the settings as well as the label

ChartMogul documents optional settings that can exclude contraction from the numerator or exclude reactivations from net retention. It also describes special handling when a report is filtered by plan. If using that service, record the actual settings; do not assume every displayed retention figure uses the unmodified formula.

For another provider, check its own definition and configuration. The fictional example uses the standard formulas stated above, a completed period and no plan filters. It is not a demonstration that every reporting tool treats all subscription events identically.

MaryChuks.com’s guide to comparing customers at the same stage addresses another comparison problem: the time customers have had to remain active. Keep that timing question alongside the revenue-definition question when preparing a real review.

A useful leadership update can celebrate expansion while making losses visible. Define the original group, distinguish gross from net retention and explain where new business enters the growth story. Then use customer evidence to decide what the business should do next.

Sources

Primary documentation checked on 31 October 2026. All customers, amounts and outcomes are fictional; arithmetic was checked locally. No commercial performance, cash receipt or customer motivation is inferred from the example.

• ChartMogul: Report — Gross MRR Retention — updated 22 October 2026.
• ChartMogul: Report — Net MRR Retention — updated 22 October 2026.