Leading Indicator

When revenue and churn were the only numbers anyone watched, every meeting was a post-mortem. The first week of April brought the report that March churn had risen to 7 percent, everyone started looking for a reason, and the customers behind that 7 percent had already deleted their accounts and left. Even a correct answer arrived too late to use that month.

So the meetings ended the same way each time. Someone asked why it went up, someone answered that it had probably been building for months, and the following quarter produced the same exchange in the same room. The numbers were accurate, they arrived late, and nothing could be changed with a number that arrives late.

A leading indicator is a number chosen to close that gap. Instead of measuring the result more precisely, it picks something earlier in the process that produces the result. It trades a little certainty for time. If the customers who left in March had been logging in less since January, a team watching login frequency in January had February to do something about it.

Walk backwards along the chain that produces a result, pick one link that moves earlier than the rest, and measure that link instead of the result.

Take paid conversion. A team signs up, creates documents, several people start editing the same document until the product has a place in their work, and months later they hit the free limit and pay. Rather than waiting for the payment at the far end, measure the share of teams that create three or more documents in their first week and you hear the same story three months sooner.

For the number you picked to do the job, three things have to hold.

1. It has to move before the result does. A number that moves in the same month is closer to a restatement of the result than a warning about it.
2. It has to be connected to the result. Line the history up with a time offset and check whether months where this number rose were followed three months later by a result that also rose.
3. It has to be observed often. A number that appears once a quarter can lead the result and still leave no room to respond.

It works like a barometer against rain on the window. Falling pressure doesn't guarantee rain, but it beats reaching for an umbrella after you are already wet. A leading indicator gives up some certainty in the same way and gets time back.

The second condition is the one that breaks most often. Plausible-looking numbers frequently have no relationship to the result at all, and an apparent relationship can just be two things moving together for a third reason. When the connection needs proving, run an A/B test that deliberately raises the number and check months later whether the result follows. A number that reaches the dashboard without that check is hard to tell apart from a vanity metric.

The first place is the seat beside the headline number on a weekly dashboard. A single north star metric sits in the middle and two or three earlier-moving numbers sit to its left. Past five of them, reading what moved this week eats the meeting and it stops being clear which one to follow.

The second is deciding experiments. A change whose result lands three months out cannot keep an experiment open that long. An experiment on the signup flow gets called on first-week document creation instead of paid conversion, and whether that call was right gets revisited a quarter later.

The third is breaking goals down. A quarterly goal of reducing churn tells a team nothing about what to do on Monday. Restated as the share of new customers who get five or more colleagues in during the first month, the work to do falls out of it.

Different roles pick different links. Marketing watches the share of visitors who take a resource rather than the visitor count. Sales watches first meetings booked per week rather than the quarter's contract value. Customer success watches how many people at an account logged in over the past two weeks rather than whether the account renews. All three are earlier links on a chain that ends months later.

"Churn hasn't moved yet, but weekly active teams have been dropping for three weeks."

A number that moves ahead of the result has turned down, so it is worth looking now.

"Nobody has checked whether this one actually connects to the result."

It moves early, but the link to the outcome was never verified.

"We can't wait until the end of the quarter, so let's call it on first-week activation."

Use the earlier link as the decision criterion for the experiment instead of the final result.

At Pagedeck, Mia keeps three rows on the Monday dashboard. Weekly collaborative documents sits in the middle, with first-week document creation and two-week invite rate beside it. When one of the three declines for more than two weeks, it becomes the first item in that week's meeting.

Last quarter, first-week document creation fell from 62 percent to 54. Mia traced the drop and found it lined up with a change to the screen shown right after signup. The rate recovered once the change was reverted, and paid conversion three months later held steady. Watching churn alone, that signal would not have arrived until after the summer.

The most common mistake is treating anything that moves early as a leading indicator. The name points at timing, so the assumption follows naturally. But moving first is only one of the conditions, and without the check that it connects to the result, what you have is a number that wobbles every week. Weekly visitors move before revenue, and in plenty of products a 20 percent rise in visitors leaves revenue exactly where it was. Put one up unverified and the team spends months pushing on it, then finds at quarter end that the result never moved and the wrong link was being held.

Believing that leading indicators are enough on their own is the other one. Fast numbers make the slow ones look unnecessary. But the link between the early number and the result snaps when the product changes — raise the free limit and teams that created plenty of documents in week one stop converting. Without lining up the lagging side every quarter, a team keeps running on a connection that already broke.