Output Metric

For a long time goals were written as a single result number. "Raise retention five points this quarter" was the standard shape. Retention is the share of customers paying this month who are still paying next month. It reads like a reasonable goal, and yet the person who receives it has no screen to open on Monday morning. Nobody types a number directly into the cell labeled retention.

Other teams leaned the opposite way and reported only activity. Twelve onboarding emails sent, eight help articles rewritten, thirty at-risk accounts contacted. Every line was true work, but three months of that reporting left retention exactly where it started, and nobody could say which of the three had mattered.

So quarterly reviews kept jamming in the same spot. The founder would ask what happened to retention, the team would read out its activities, and when told that activities are not a result, the team would ask how retention is supposed to go up. Both sides had a point. The name came out of splitting numbers into two kinds at that impasse: the result side is output, the side a team can move directly is input. Deciding which side a number sits on requires its metric definition first — the written rule for what gets counted and how. Without that rule you can't even test whether the number reacts to anything the team does.

Fix one counting rule and one reporting cadence, then repeat the same arithmetic every period. No single formula belongs to the category; each output carries its own. Monthly retention at Pagedeck divides the teams that paid last month and paid again this month by the teams that paid last month. If 294 of last month's 320 teams paid again, that's 91.9 percent. Put monthly revenue in the same slot and the formula becomes paying teams times average spend per team.

Three decisions come first.

1. Write down, in one line, the decision this number will drive. For retention that might be whether to add customer success headcount next quarter. Skip this line and the other two decisions have nothing to be judged against.
2. Nail the counting rule. Whether free teams count, how annual contracts are handled in the months when no charge lands, whether a team one day late on payment counts as churned — all of it gets written down. Leave it unwritten and the same metric name grows as many versions as there are people reporting it.
3. Pick the cadence. Read an output too often and it swings for reasons that have nothing to do with the team. Look at retention weekly and two contract renewals landing in the same week move the number several points.

Because a lag sits between the action and the result, an output on its own can't tell you what to do this week. The usual move is to break it apart and find something movable underneath. A metric tree is that structure — one large number split by multiplication and addition into the parts hanging below it. Monthly revenue becomes paying teams times average spend per team, and paying teams splits again into new plus retained plus reactivated. Keep going down and you reach something like onboarding completion rate for new teams, which somebody can work on this week. That is an input metric.

No absolute threshold exists. Ninety-two percent monthly retention is unremarkable for a product billed month to month and alarming for one sold on three-year contracts. So the judgment rests on three conditions rather than on the value itself.

The first is the shape of the trend. Readings of 88, 89 and 91 move in one direction and can carry next quarter's plan. Readings of 91, 86 and 93 average to 90, and that 90 guarantees nothing. Swings that wide usually come from a small sample or from a counting rule that changed halfway through, so check both before interpreting the number.

The second is whether the link to an input has been verified. A usable output metric is paired with at least one input, with a record of the output moving when that input moved. Tripling onboarding emails for a month and watching week-four retention go from 61 to 68 percent establishes the link. If retention held flat through the same period, that input comes off the list and the search continues.

The third is how long the reaction takes. A number whose lag runs past a quarter makes a poor quarterly target. When the effect of this quarter's work surfaces six months later, this quarter's score gets filled in by whoever held the job before. Numbers like that belong on a watch list, and the target moves down to something with a shorter lag. Among several outputs, the one promoted to represent the whole company has its own name: the north star metric.

"That's an output, so we can't make it the sprint goal. Let's find an input we can actually move."

The number won't respond to what the team does this week, so a different number should carry the execution-level goal.

"We beat the output target last quarter but every input metric came in under plan."

The good result may not have come from anything the team did, so nobody should expect a repeat next quarter.

"Retention is an output, so let's review it monthly and watch onboarding completion in the weekly."

Split which numbers belong to which meeting cadence, and stop staring at a slow-moving number every week.

The distinction earns its keep at Pagedeck's quarterly goal-setting session. When Grace proposes a three-point improvement in retention, Mia opens the metric tree and shows what hangs underneath it: onboarding completion rate, time to first document created, share of accounts where someone invited a teammate. Two of those become team goals for the quarter, while retention itself is checked monthly for direction only.

The most common mistake is reading outputs as the more important numbers. They are the ones that travel upward in reporting, so the impression forms easily. But the two kinds sit side by side rather than in a hierarchy: outputs answer whether the direction is right, inputs decide what happens this week. Rank them and leave only outputs on the dashboard, and the weekly meeting turns into reading last week's numbers aloud with nothing to change.

The second mistake is assuming that grading a team on outputs makes accountability clearer. Judging results feels fairer than judging activity, so the reasoning is natural. The problem is that outputs absorb factors the team cannot touch. A competitor cuts prices, a large customer disappears in a reorg, and the number moves on its own. A team graded on something it can't control starts improving the number instead of the business — dropping accounts that are certain to churn out of the denominator, or pushing a renewal date into the next quarter. That is Goodhart's law in plain view: the moment a number becomes a target, it stops measuring.