KPI Picker

Leading and lagging measures

Most KPI lists mix two kinds of number without saying so. A lagging measure is the result itself: revenue, churn, profit, customers kept. It tells you whether you won, but only after the fact, when it is too late to change the outcome. A leading measure moves earlier and predicts that result.

The idea is not new. When Robert Kaplan and David Norton introduced the balanced scorecard in Harvard Business Review in 1992, their starting point was that financial measures on their own can give misleading signals about improvement and innovation¹. The book The 4 Disciplines of Execution (4DX) gives the sharpest test for a lead measure. It has to pass two checks: it predicts the goal, and the team can influence it². A number that fails the first is busywork. A number that fails the second is just another result to watch.

A weak KPI set and a stronger one

The team, goal and numbers below are invented for illustration. A customer success team at a software company wants to cut the share of new customers who cancel in their first 90 days.

The weak set counts effort. Calls and emails are easy to raise and may have nothing to do with whether customers stay. Each measure on the right is tied to the one result the team is chasing.

Test whether a lead measure really predicts results

A leading measure is a guess about cause and effect until you check it. Accounting researchers Christopher Ittner and David Larcker tested this kind of guess in 1998, asking whether customer satisfaction scores predicted later financial results³. In Harvard Business Review in 2003, they warned that many companies track nonfinancial measures without ever testing whether those measures drive the results they care about⁴.

A team can run a simple version of that test without a statistician:

• Look back first. Pull the last 6 to 12 months. In the example above, compare new accounts that finished setup in 14 days with those that did not. If both groups cancel at the same rate, the measure does not predict the result.
• Look forward after a quarter. Once the team has pushed the lead measure up, check whether the lagging measure followed. If it did not, change the lead measure rather than working harder on it.
• Watch for a shared cause. Customers who were keen from day one may both finish setup and stay, whatever the team does. Trying the change with one group first, then comparing, is stronger evidence than a pattern in old data.

How many KPIs is too many

The research for this page found no study that sets a right number of KPIs. What studies do show is that people do not use long lists evenly. In an experiment reported in The Accounting Review in 2000, people judging business units paid attention to the measures all units shared and largely ignored the measures unique to each unit⁵. At one financial firm studied by Ittner, Larcker and Marshall Meyer in 2003, managers put most of the bonus weight on financial measures and paid little attention to measures that predicted future results⁶. Extra measures on a list do not mean extra attention.

One measure is not the answer either. In two experiments reported in 2012, paying people on a measure made them more likely to treat the measure as the strategy itself, and the effect was stronger with one measure than with several⁷.

A workable shape for each goal: one lagging measure, two or three leading measures that pass the 4DX test, and one or two health checks that are watched but not targeted. 4DX also advises no team work on more than two goals at once². This tool returns five or six KPIs for an area. Keep the ones tied to your goal, label each as leading or lagging, and drop the rest.

When the number becomes the goal

Economist Charles Goodhart observed in 1975 that a statistical pattern tends to break down once pressure is put on it for control⁸. He was writing about monetary policy in the UK, not business KPIs. The line most people quote, "when a measure becomes a target, it ceases to be a good measure," is a later rewording by anthropologist Marilyn Strathern, in a 1997 paper about audits of British universities⁹.

The business version has a name: surrogation, treating the measure as if it were the strategy. Michael Harris and Bill Tayler, writing in Harvard Business Review in 2019, describe Wells Fargo staff opening 3.5 million accounts without customer consent while chasing a cross-selling measure¹⁰. Their advice: involve the people doing the work in shaping the strategy, use several measures, and avoid tying pay to a single one¹⁰.

Guards that fit on any KPI list:

• Pair each target with a check. Tickets closed goes with tickets reopened. Deals won goes with discount given. Setup finished goes with support tickets.
• Write the goal in words next to the number. "Customers get value in their first two weeks" is harder to game than "setup rate."
• Read a few real cases. Once a month, look at five customers behind the number and ask whether it still describes what happened to them.

Turning the list into measures a team runs

A KPI with no owner, no starting point and no review date is a label. For each one you keep:

• Name one owner. A person, not a team, who reports the number and raises it when it drifts.
• Record the baseline. The formula the tool gives only helps once you know today's value.
• Match the cadence to how fast it moves. Leading measures are worth a look every week. Lagging ones often change too slowly to read more than monthly.
• Set a date to test it. Put the look-forward check from above on the calendar for the end of the quarter.

To build lead measures into a weekly rhythm, see the 4 Disciplines of Execution. To balance financial results against customer, process and learning measures across a whole company, see the Balanced Scorecard. To turn a KPI you want to move into a time-boxed goal, try the OKR generator.