Metrics5 min read

Leading and Lagging Indicators

Revenue tells you what happened. You need something that tells you what is about to.

It is week ten of the quarter and the revenue number is short. The team has been working hard the whole time. The question in the room is what went wrong in weeks one to nine, and nobody can answer it, because the only metric anyone was watching was the one that takes nine weeks to move.

Lagging indicators confirm outcomes. Revenue, churn: The rate at which customers stop paying or stop using the product., retention at day 90, annual contract value. They are trustworthy and they are late. By the time they move, the decisions that caused them are months old and the people who made them have moved on.

A leading indicator, checked against the outcome

The candidate moves first and the outcome follows several weeks later, which is what makes it worth steering by. The check is this shape in your own history. Plenty of plausible leading indicators never produce it, and steering by one of those is expensive.

Leading indicators move first and predict the lagging ones. They are less trustworthy on their own, because a correlation you found once may not hold, and they are the only metrics you can actually steer with.

Lagging

Quarterly revenue
Annual churn rate
Day 90 retention
Net revenue retention
Confirms. Cannot be steered inside a quarter.

Leading

Trials that complete first key action in 24 hours
Accounts with a second active user in week one
Support tickets per hundred active accounts
Share of new accounts importing existing data
Steers. Must be validated against the lagging one.

The trap is adopting a leading indicator: Something measurable now that tends to predict a result you care about later. because it sounds sensible rather than because you have checked it. A leading indicator that does not actually predict anything is worse than no metric, because the team will optimise it in good faith for two quarters before anyone notices the lagging number never followed.

Worked example

Hypothetical: Halverson, a B2B contracts tool

The team suspects that uploading five or more existing contracts in week one predicts retention. They check against last year's cohort: A group of users bucketed by when they joined, tracked over time rather than blended together.. Say among accounts that hit five uploads, about 70 percent were still paying at month six, versus about 35 percent of accounts that did not. That is a large enough gap to steer by. But note what the check cannot tell you: whether uploading causes retention, or whether serious buyers both upload and stay for the same underlying reason. Both stories fit the data. That matters, because if it is the second story, pushing everyone toward five uploads with nudges and incentives will move the leading indicator and leave retention exactly where it was. The only way to separate the two is to change the onboarding for a random half of new accounts and watch whether retention follows.

Building the pair

Name the lagging outcome: The thing you actually care about and cannot move directly. Be precise about the horizon: retention at 90 days behaves differently from retention at a year.
Find candidate behaviours: Early actions that differ between accounts that reached the outcome and accounts that did not. Look in the first week or two, where you still have time to intervene.
Check the gap: Compare the outcome rate for accounts that did the behaviour against those that did not. A small gap means a weak signal, however intuitive the behaviour feels.
Test causation: Change something to drive the behaviour for a random subset, then watch the lagging metric. Until you do this you have a correlation you are treating as a lever.
Re-validate: The relationship decays as the product and the audience change. Re-check it at least every two quarters.

A good leading indicator also has to be one your team can influence this week. Accounts with a second active user in week one is leading and actionable. Market sentiment is leading and useless, because no work you schedule changes it. If a candidate passes the prediction test but nobody can name a project that would move it, it is a forecast, not a lever, and it belongs on a dashboard rather than in a goal.

Watch the horizon mismatch as well. If your leading indicator moves in days and the lagging one takes two quarters to respond, you will get several rounds of encouraging news before the first piece of real evidence arrives. Say this out loud when you set the goal, so that a rising leading indicator in week three is reported as early and unconfirmed rather than as a win.

Report both, always together. A leading indicator without its lagging partner invites gaming. A lagging indicator: A result that confirms what happened but arrives too late to steer by. without its leading partner invites a room full of people explaining the past.

Quick check

A behaviour strongly correlates with retention. Why is that not enough to target it?

The takeaway

Leading indicators let you steer, but only once you have checked that they actually predict the lagging outcome you care about.

Try this tomorrow

Pick your team's main leading metric and pull the retention rate for accounts that hit it versus those that did not. If the gap is small, replace the metric.

Answer the check above, then bank the day.

Where this comes from

Measure What Matters, John Doerr
Radical Focus, Christina Wodtke
Hacking Growth, Sean Ellis and Morgan Brown

Metrics is one of six tracks. These lessons summarise and build on the work above, they do not reproduce it. Buy the books, they are better.