Customer success metrics that actually matter
A small CS team doesn’t have a metrics problem, it has an attention problem. You can track forty numbers or you can act on five. This is the five, why each one earns its place, and the popular ones you can safely skip.
NRR and GRR, the two numbers the business runs on
Start with the money. Net revenue retention (NRR) tells you whether your existing base is growing: renewals plus expansion, minus churn and downgrades. Above 100% and the business would grow with zero new sales. Gross revenue retention (GRR) strips expansion out and shows only what you kept, so it can never exceed 100%. It is the honest measure of the leak.
You need both because expansion can hide churn. A book with strong NRR and weak GRR is losing customers and papering over it with a few big upsells, which works right up until the upsells slow down. Report the pair, watch the trend, and treat a sliding GRR as the earlier alarm.
Logo churn vs revenue churn, and why they diverge
Logo churn counts customers lost. Revenue churn counts dollars lost. On a book where every account pays the same, they move together. On a real book they diverge, and the divergence is the story. Ten small accounts leaving might be 2% of revenue and a product-fit signal at the low end of your market. One large account leaving might be 15% of revenue and a relationship failure you should have seen coming.
Track both, but weight your attention by dollars. A lean team’s save effort should go where the revenue is, which is why a flat “accounts at risk” count is less useful than the ARR at stake behind it.
Health score, the one leading indicator
Everything above is a lagging metric. By the time churn shows up in NRR, the customer is gone and the quarter is closed. A customer health score is the one number on this list that looks forward: it reads the signals that precede a renewal decision, usage, reply times, ticket tone, billing behavior, and condenses them into a per-account read you can act on while the outcome is still open.
The catch is that a health score is only as good as its inputs. A score built on “logged in this month, yes or no” will be green right up to the cancellation email. A useful score has to read what customers actually do and say, across systems, and it has to name its evidence so you can trust a red flag enough to act on it.
Time-to-value, the metric that predicts year one
Time-to-value measures how long a new customer takes to get the outcome they bought you for. Not onboarding checklist completion, the actual outcome: the first report shipped, the first workflow live, the first result their boss can see. It is commonly observed that customers who reach value fast renew, and customers who stall in onboarding churn at the first renewal, so this number is effectively an early read on next year’s GRR.
For a small team it is also the most fixable metric on the list. You control the onboarding path directly, and shortening it pays back on every future customer. Define the value moment for your product, timestamp it per account, and watch the median.
Vanity metrics you can drop without guilt
NPS on its own.Sentiment surveys answer “how do you feel today,” not “will you renew.” A customer can rate you a 9 and churn three months later because the champion left or the budget moved. NPS is fine as a supplement to behavioral data; as a headline metric it mostly produces false comfort.
QBR counts. Meetings held is an activity number, not an outcome number. A quarterly review that changes nothing still increments the counter. If you track anything here, track what came out of the conversation, not that it happened.
Ticket volume without context. Raw ticket counts cut both ways. Rising tickets can mean a struggling customer or a deeply engaged one. Falling tickets can mean things are fine, or that the customer stopped trying. The signal is in the tone, the type, and the change, never the count alone.
Five numbers, one page
NRR, GRR, churn (logos and dollars), health score distribution, time-to-value. One page, same cadence, every week or every month depending on the size of your book. The discipline isn’t in the dashboard, it’s in the refusal: every metric you add divides the attention you have for the ones that move revenue.
When someone asks for a metric that isn’t on the page, ask what decision it would change. If the answer is “none, but it would be good to know,” it goes in a report nobody has to read, not on the page.
The metrics are the output. The reading is the work.
The hard part isn’t computing five numbers, it’s the daily reading of email, tickets, usage and billing that makes the health score mean something. Keply does that reading: it scores every account from real signals, flags the ARR at stake, and drafts and sends the save with your approval by default. Connect your tools or upload a spreadsheet, live in a day, flat $299/month. See how it works.
CS metrics, FAQ
- What are the most important customer success metrics?
- For a small team: net revenue retention (NRR), gross revenue retention (GRR), logo churn alongside revenue churn, a health score as your leading indicator, and time-to-value. Those five cover the money, the leak, and the early warning. Everything else is optional until you have the headcount to act on it.
- What's the difference between logo churn and revenue churn?
- Logo churn counts customers lost; revenue churn counts dollars lost. They diverge whenever your accounts differ in size. Losing ten small accounts can be 2% of revenue, losing one large account can be 15%. Track both, because each one hides a failure mode the other exposes.
- Is NPS a useful customer success metric?
- On its own, not really. NPS measures stated sentiment at one moment, and a customer can score you a 9 and still churn when the champion leaves or the budget gets cut. It's a useful supplement to behavioral signals like usage and renewals, but a poor substitute for them.
- How many metrics should a small CS team track?
- Around five, on one page, reviewed at the same cadence. The limit isn't the dashboard, it's attention. A team of one or two can act on five numbers. A twenty-metric scorecard mostly produces reporting work and no different decisions.
See the revenue at risk on your own book.
Connect your stack or drop in a spreadsheet. Live in a day.
Start freeFlat $299/mo. First month free. 30-day money-back. Cancel anytime.