Lagging Indicator

There was a time when "looking at the metrics" was one undivided activity. Revenue and churn went into the monthly report, the team discussed what to do about them, and the numbers were accurate. The conclusions kept failing to turn into anything.

Picture the meeting where March churn lands in your inbox during the second week of April. Someone asks what caused it. The meeting decides to rewrite the onboarding emails. But the customers those emails could have saved left weeks ago. The number under discussion and the action coming out of the discussion were pointed at two different sets of people.

So metrics got re-sorted by when they move rather than by what they measure. Numbers that finalize only after the outcome is complete became lagging indicators, and numbers that move ahead of the outcome and hint at what's coming became leading indicators. Once both names existed, meetings changed order. Lagging indicators close the books; leading indicators decide what happens this week.

A lagging indicator sums up the facts of a period that has already closed and confirms, after the fact, whether the judgment made during it held up.

It has three properties. First, the value settles. March revenue locks in around mid-April once refunds and failed payments clear, and it does not move after that. Second, there is a lag between cause and value. Almost nothing that shows up as March churn was caused in March; it usually traces back one or two quarters. Third, it is unrecoverable. By the time you read the number, the events that produced it are over.

Monthly revenue, quarterly churn, and free-to-paid conversion are the standard examples. Reading them takes three steps.

1. Close the period. A figure read before the accounting settles wobbles from day to day. March churn checked on March 20 is missing every account whose billing date falls at the end of the month.
2. Pair it with an earlier period's behavior. Comparing March churn to what the team did in March turns up almost nothing. Comparing it to what January's cohort did in its first week is where the connection appears.
3. Line it up against a leading indicator. If January onboarding completion fell from 62 percent to 48 percent and March churn rose from 4 percent to 7 percent, check whether that link actually holds in the data rather than assuming it.

The reason to keep watching a number that arrives late is that it is certain. A leading indicator always has a layer of inference in it. The observation that teams creating three documents in week one tend to stick around may or may not keep holding, and only the outcome settles that. The lagging indicator is the outcome. If a leading indicator is a prediction, a lagging indicator is the graded answer sheet.

So the two get used as a pair. Move on the leading indicator, confirm on the lagging one. Watch only lagging indicators and you are always late; watch only leading ones and you end up diligently raising the wrong thing.

This term surfaces in the meeting that closes out a period, not the one that decides what to do this week.

The first of those is reporting. Board decks and investor updates are made almost entirely of lagging indicators — quarterly revenue, churn, customer count — because a number has to be closed and final before it goes outside the company.

The second is validating a leading indicator. Whether the number you picked as leading actually leads is knowable only once the lagging indicator arrives. If a quarter of work raised onboarding completion by 15 percentage points and churn two quarters later is unchanged, that number was not a leading indicator. Skip this check and a team can spend several quarters raising a figure that connects to nothing.

The third is setting targets. Quarterly targets get set on lagging indicators while weekly reviews run on leading ones. Target a leading indicator directly and the team drifts toward raising that number alone; run a weekly review on a lagging indicator and you stare at the same screen for four months without a decision to make.

On a dashboard, put the north star metric in the middle, two or three leading indicators to its left, and two or three lagging indicators to its right. Keep vanity metrics off the board entirely, and lay a counter metric along the bottom to watch what might break while the target goes up.

"That's a lagging indicator, so there's nothing to act on. Let's look at the earlier numbers."

The period is closed, so this isn't something to hold a response meeting about.

"We need to read this quarter's churn next to January's onboarding completion."

One lagging indicator alone won't locate the cause, so pair it with a leading indicator offset by the right lag.

"Set the target on lagging indicators and run the weekly check on leading ones."

Separate the standard for settling results from the standard for running the work.

The split took hold at Pagedeck during an April metrics review. March churn came up on screen at 7 percent, and the first reaction was to roll back a feature change that had shipped in March. Mia put up a version of the table cut by signup month instead. Most of the churned accounts were teams that had signed up in January, and January was the month the signup flow had been reworked.

The review screen has been split in two ever since. On the left sit this week's onboarding completion and first-week document counts. On the right sit last quarter's churn and conversion. The left side gets acted on the same week; the right side is used to grade decisions made two quarters ago.

The most common mistake is concluding that a lagging indicator is useless because it's late. Hearing "look at the earlier numbers" enough times tends to settle into that. But this is the only confirmed fact the team has, and there is nowhere else to check whether a leading indicator was right. Throw it out and leading indicators start grading themselves. A team can report two quarters of rising onboarding completion as a win while churn sits exactly where it was, and nobody catches it.

The second is treating the character as a fixed label on each metric — revenue is lagging, active users are leading, memorized once and reused. In practice the slot changes with what you pair the number against. Monthly revenue leads annual revenue and lags that month's sales activity. The label belongs to the gap between two numbers, not to one number on its own. Memorize it as a fixed tag and you end up arguing about classification in meetings while never doing the actual work, which is measuring the lag.