Customer Success Metrics: Definitions, Formulas and Benchmarks

Churn rate: the leakage measure

Customer churn rate is the share of customers lost in a period. Revenue churn is the share of recurring revenue lost. They answer different questions, and reporting one while implying the other is the most common reporting error in early-stage customer success teams.

Customer churn rate = customers lost in period ÷ customers at start of period × 100. A team that starts a quarter with 200 accounts and loses 8 has a 4 percent quarterly churn rate.

Revenue churn rate = recurring revenue lost ÷ recurring revenue at start of period × 100. Losing $40,000 of a $1,000,000 starting base is 4 percent revenue churn — the same headline number, but a very different story if the eight accounts lost were your smallest or your largest.

Always state the period and whether the figure is customer or revenue based. Annualizing a monthly rate by multiplying by twelve overstates the loss; compound it instead.

Gross and net revenue retention

Gross revenue retention (GRR) measures how much recurring revenue you keep from the existing base before any expansion. It can never exceed 100 percent, which makes it the honest measure of leakage.

GRR = (starting recurring revenue − downgrades − churn) ÷ starting recurring revenue × 100. From a $1,000,000 base with $40,000 churned and $20,000 downgraded, GRR is 94 percent.

Net revenue retention (NRR) adds expansion back in. NRR = (starting recurring revenue − downgrades − churn + expansion) ÷ starting recurring revenue × 100. With $90,000 of expansion on the same base, NRR is 103 percent.

Read them as a pair. A healthy NRR sitting on a weak GRR means expansion from a few accounts is masking steady attrition elsewhere — a pattern that holds until the expanding accounts stop expanding.

Time to first value

Time to first value measures the days between contract start and the customer's first measured outcome. It is the earliest reliable predictor of retention available to a customer success team, and the one most often left unmeasured because the outcome was never defined.

Define the outcome per use case before onboarding starts, then measure the median rather than the average — a handful of stalled implementations will distort an average and hide an otherwise healthy process.

Track the distribution too. A median of 21 days with a long tail past 90 tells you to fix the tail, and the tail is usually a dependency on the customer's side that nobody owned.

Adoption breadth and depth

Adoption is two measures, not one. Breadth is the share of licensed users active in the period. Depth is the share of the core workflow they actually use.

An account with 80 percent breadth and shallow depth is using your product as a lightweight tool and will be easy to replace. An account with deep usage among three of forty users is a single-champion risk: when that person changes role, the renewal becomes a conversation with strangers.

Health score: components over a single number

A health score is a weighted composite, and its value depends entirely on whether the components are visible. Report the distribution across the book and the component driving each low score.

A workable starting model: usage against expected pattern (30 percent), adoption breadth (20 percent), support experience including repeat contacts (15 percent), sponsor engagement (15 percent), outcome progress against the success plan (20 percent).

Validate it backward every quarter against accounts that actually churned. A model that was green ninety days before a loss is missing a signal, and finding that signal is worth more than refining the weights.

CSM KPIs that do not distort behavior

Individual KPIs change behavior, including in ways you did not intend. Measuring meetings held produces meetings. Measuring renewals alone pushes risk reporting underground, because nobody volunteers bad news about their own number.

A balanced set: gross retention on the assigned book, net revenue retention, time to first value for new accounts, on-time completion of the review cadence, and risk identification accuracy — how often a flagged risk turned out to be real and how often a churn event was never flagged.

That last measure is the one that keeps a health score honest. It rewards the CSM who raises a hard problem early instead of the one whose dashboard stays green until the cancellation email.

Reporting cadence and the CTR of your own metrics

Weekly, review the risk list and the onboarding pipeline. Monthly, review themes, adoption trends and the review cadence. Quarterly, review retention, NRR, the health model's accuracy and the segment-level picture.

Cut any metric that has not changed a decision in two quarters. Reporting effort spent on unused numbers is the quietest cost in a customer success operation.

Frequently asked questions

What is the formula for churn rate?

Customer churn rate = customers lost during the period ÷ customers at the start of the period × 100. Revenue churn rate uses recurring revenue instead of customer counts. Always state the period and which basis you used, because the two can tell opposite stories about the same quarter.

What is the difference between gross retention and net retention?

Gross revenue retention excludes expansion and therefore caps at 100 percent — it measures leakage from churn and downgrades. Net revenue retention includes expansion and can exceed 100 percent. Read them together: strong NRR over weak GRR means expansion is hiding attrition.

What is a good net revenue retention benchmark for B2B SaaS?

Published US B2B SaaS benchmarks commonly place solid performance above 100 percent, with enterprise-focused companies reporting higher figures than SMB-focused ones because expansion is easier in larger accounts. Compare against your own segment and trend rather than a single headline number, and verify any benchmark against the current year's published source before quoting it.

Which customer success metrics should a small team track first?

Gross revenue retention, net revenue retention and time to first value. Those three cover leakage, growth and delivery speed. Add a four-component health score once you have enough accounts for the pattern to mean something.