Every indicator is a rear-view mirror.

Reacting one cycle late, every time

Management reacts to the past, one reporting cycle late, every time.

The indicator turns in month four. By the time it is reported and discussed it is month five, and whatever caused it began in month three. The response is therefore always aimed at something that has already finished happening, which is why recovery plans so often address the previous problem.

What a leading indicator looks like

The distinguishing feature is that it measures the conditions for work rather than the work.

Constraint closure rate: how many of the things blocking next month's activities were cleared this month, from Week 9. Approval turnaround: how long submittals are taking now, from Week 15. Ready workfronts: how many are genuinely available against how many the look-ahead needs. Drawing release against the construction sequence rather than against the engineering plan.

None of these is progress. All of them predict it, and all of them are already being collected somewhere on your project for another purpose

.

The one that gets gamed

An indicator that somebody is judged on becomes a target, and a target becomes a number that is managed rather than measured.

The recognisable symptom is a figure that always lands just under the threshold. Progress that is consistently one point below plan, month after month, isn't a measurement of anything. It is a narrative, and it means the estimating

discussed in Week 8 is being tuned to the reporting rather than to the wall.

Which is an argument for indicators nobody is judged on. Constraint closure works partly because it isn't yet a performance metric anywhere.

How many is too many

Packs grow. Every question anybody has ever asked leaves an indicator behind, and removing one feels like admitting it was never useful.

The test isn't whether an indicator is interesting. It is whether anybody has changed a decision because of it in the last six months. Most packs contain a handful that pass and a long tail that nobody has looked at since the month they were added.

The tail isn't harmless. It buries the few that matter, and it makes the pack long enough that the reader skims, which returns us to the failure in Week 26. A page with six indicators that are read beats a page with twenty that aren't.

Measuring the error before correcting it

If the pack is yours to design, then for every lagging indicator in it, add one leading indicator that predicts it, and keep the pack small enough that both get read.

Inherited: don't propose replacing the pack. Add two leading indicators to the existing one and let them run for three months alongside. If they turn before the lagging ones do, the case makes itself. If they don't, you have learned something about your project rather than lost an argument.

System design

For every indicator that describes a finished period, one that describes conditions now. All four of the second kind are already being collected for another purpose.

Record	Produced by	Required quality	Verified against	Feeds
Progress against plan	Project controls	Usable as is — and describes a finished period	The baseline	Performance review
Schedule and cost indices	Project controls	Usable as is — and describes a finished period	Cost report	Performance review
Constraint closure rate	Project controls	Count of closures with evidence, not verbal	Constraint log	Whether next month starts
Approval turnaround	Project controls	Observed, rolling, not contractual	Submittal log	Engineering and procurement dates
Ready workfronts	Construction	Passing all six tests, against what the look-ahead needs	Workfront register	Achievable look-ahead

Add two and run them alongside the existing pack for three months rather than proposing a replacement. If they turn before the lagging ones do, the case makes itself.

Practical insight

List the indicators in your monthly pack and mark each one: does it describe a period that has ended, or conditions that exist now?

Most packs come out entirely in the first category. Then pick one thing you already collect that predicts next month — open constraints, submittals outstanding, workfronts ready — and put it on the same page. It costs one line and it is the only line on the page about the future.

Key takeaways

• Standard indices describe a period that has finished. They are necessary and they predict nothing.
• Management therefore responds a cycle late, to something that has already stopped happening.
• Leading indicators measure conditions for work: constraints closed, turnaround, ready workfronts, drawing release.
• All of them are already collected somewhere for another purpose.
• A figure that always lands just under target is a narrative, not a measurement.
• Indicators nobody is judged on stay honest longer.
• On an inherited pack, add two and run them alongside rather than proposing a replacement.

Records born here. KPI page with both kinds marked · the leading indicator series, kept from the start.

What is coming next

Indicators point at things. What records them, and what closes them, is the next question.

Next week: minutes, actions and closure — and why the fourth step is the one that gets skipped.