Customer Success Metrics

Sales ends a quarter with signed contracts. Marketing has sign-ups, engineering has shipped features. Customer success is the odd one out, because when it goes well the result is not an event but the absence of one. An account someone spent three months saving looks exactly like an account that was never in danger. A cancellation that got prevented never happened, so nobody sees it.

Budget conversations used to stall in the same spot every time. Asked for two more people next quarter, the manager would get back a simple question: what do those two produce? The honest answer was that customers seemed happy, and happiness has no unit, so the conversation ended there. In the same meeting sales read out twelve closed deals and marketing read out eight thousand sign-ups. The function looked like pure cost inside the company, not because the work was bad but because there was no column to write the result in.

The first column anyone built was the count of customers who left. A hundred accounts at the start of the quarter, five gone by the end, five percent. That single line stayed in use for a long time, and it erases about half of what actually happened in the quarter. If the five that left were small teams paying a hundred dollars a month, and a handful of the remaining ninety-five added seats worth twenty thousand dollars, the account count shrank while the money grew. Turn it around: if the five that left were the largest accounts on the books, five percent of the customers can be forty percent of the revenue. One line cannot tell those two quarters apart.

Once the numbers were split up, a different sentence started appearing in the same meeting. How much money leaked out of the existing base, how much got added to it, how many days a new account waited before it got anything useful out of the product — each of those had its own row. A request for more headcount could now arrive attached to something specific, such as the fact that the wait for new accounts had been growing for three straight weeks.

That is why one measure turned into several. They fall into two groups. One group reads money out of the billing records, the other reads behavior out of the product's own usage logs. The first group describes what already happened with real precision. The second describes what has not happened yet, usually a couple of months before it lands in the billing records. Only after teams started reading both did customer success move from reporting results at the end of a quarter to acting in the middle of one.

Five terms sit under this name, and they split by where the number comes from. Two are pulled from billing records, three from product usage.

The two billing measures ask how much of what existing customers were paying a year ago is still being paid today. Newly signed customers are deliberately excluded. Mixing them in would fuse the part sales won with the part customer success defended into a single figure, and neither team could then point to its own contribution.

1. Gross revenue retention (GRR) is the share of existing revenue that survived. Suppose those customers were paying 100,000 dollars a month a year ago, and since then cancellations took away 5,000 and downgrades another 3,000. What's left is 92,000, so GRR is 92 percent. Nothing that grew gets added back, which means this number can never exceed 100 percent and no amount of new business can hide it. It shows the exact size of the leak and nothing else.
2. Net revenue retention (NRR) runs the same calculation and then adds the growth. If the surviving customers in that example bought more seats and moved to higher plans worth another 15,000 dollars, the total reaches 107,000 and NRR is 107 percent. Because there is something to add, this figure can go past 100, and going past 100 means revenue grows even in a quarter with no new customers at all. That property is why investors tend to ask for it first.

One more habit goes with these two: fixing the group before you start counting. The set of customers is frozen as it stood twelve months ago and then followed, rather than recalculated each month against whoever happens to be on the books. Otherwise a wave of new signings quietly enlarges the denominator and the retention figure improves without a single existing customer having stayed longer.

The two are always read side by side. A company at 80 percent GRR and 105 percent NRR is growing on the surface while losing one account in five underneath, with a small number of expanding customers covering the hole. If one of those large expanding accounts leaves, both numbers collapse together.

The three usage measures describe the state customers are in right now. The billing pair only settles when a contract comes up; these recalculate every week, which is what leaves time to act before a renewal arrives.

1. Time to first value is the stretch between the signature date and the day the customer gets a real result out of the product. Pagedeck defines that moment as five people from one team editing the same document within a single week. Choosing that definition is half the metric — change the bar and the whole number moves, which makes a definition change easy to mistake for a performance change. The same reasoning is why the median is used instead of the average. One account that took six months drags an average up by itself and buries what everyone else experienced.
2. Usage depth is how many of the product's features an account actually touches. Pagedeck has twelve core features and a typical account uses three of them. A shallow account can move to a similar tool whenever it likes, so those are the ones that wobble first at renewal. An account that has built its approval process inside the product is in the opposite position: switching tools would mean rewriting how the company works, so it rarely leaves.
3. Customer effort is how much work it took the customer to get what they came for. It's usually collected right after a support conversation closes, by asking the person to rate a statement like "it was easy to resolve my issue" from one to seven. Where it parts company with satisfaction is direction. A smooth experience fades from memory; an annoying one is still there at the renewal meeting. A customer who needed four rounds of email to get one answer will rate the product well and still hesitate over the next contract.

The five run in time order. When time to first value stretches, usage depth settles shallow a few weeks later; shallow accounts hit more walls, so effort scores get worse; and an account in that state comes up for renewal and drops out of gross revenue retention. Dig into a quarter where the money numbers went bad and the answer usually sits in usage logs from two or three quarters earlier. Pagedeck watched exactly that: the cohort that signed during a slow onboarding stretch renewed at 88 percent against a normal 94, and for the two quarters in between the billing numbers looked untouched. The reverse failure is just as real — watch usage alone and you can end up with a growing crowd of happy small accounts and flat company revenue.

"Gross retention is 92 and net is 107. Revenue is growing even without new logos."

More money is being added inside the existing base than is leaking out of it, so this quarter's growth doesn't have to be explained by sales performance.

"Median time to first value went from 19 days to 34 this quarter. Those accounts renew this time next year."

Nothing is wrong with the current numbers; the warning is about the renewals twelve months out.

"Forty percent of accounts sit at usage depth one. Let's treat that group as the renewal risk list and split it up."

Accounts using a single feature are cheap to replace, so someone should be assigned to them before the contract date arrives.

Pagedeck reads these numbers together in the first week of every month. Sofia goes through the money that left through cancellations and downgrades, then the money added through extra seats, and works the two retention figures out in front of everyone. Grace always asks the same follow-up: how many accounts did the loss come from? Spread evenly across ten accounts, it points at the product or the pricing. Concentrated in one, it's a story about that one company. The response is completely different, so the total on its own settles nothing.

Mia handles the usage side in the same meeting. Last quarter the median time to first value grew by nearly two weeks, and pulling it apart showed that only accounts rolling out to more than twenty people were slow. Something in the import step was jamming for them. Theo fixed it and the median came back down the following month; the effect reached the billing numbers two quarters after that.

The most common mistake is reading these numbers as a report card for the customer success team. The name has customer success in it and the team reads them aloud, so the assumption is natural. But the biggest lever on retention is not the effort spent three months before renewal — it's the decision made on the day the deal was signed. A customer who arrived expecting the product to solve something it doesn't solve will leave a year later no matter what the team does in between. Hang the number on customer success without that distinction and the team spends its energy answering for an outcome it didn't create, while the part that actually needs fixing, which is who gets sold to, stays exactly as it was. So when retention slips, the first cut is by acquisition channel and by what those departed accounts expected at signing.

The second mistake is believing one headline number is enough. Picking a single metric per team is fashionable, and in customer success that seat usually goes to net revenue retention. Reporting does get simpler. The trouble is that this figure only settles when contracts close, which makes it a lagging measure. You learn in March that March went badly, but the behavior that produced it finished last autumn, and by the time you know, there is no customer left to win back. A team holding one lagging number ends up running a quarterly meeting about causes rather than a monthly one about actions. Doing something requires figures that can be read today from customers whose renewal hasn't arrived yet, which is what the three usage measures are for.