Leading Indicators Are a Behaviour, Not a Metric
A signal is only leading if someone is willing to act on it while it can still be denied.
When the Report Reads Green
Every experienced programme director has lived through the same unsettling sequence. The monthly report goes to the board and it reads green. Milestones met. Spend within tolerance. The risk register is tended, the issues are being worked, the plan shows the next stage gate comfortably in reach. Three months later the same programme is unmistakably in trouble — a slipped gate, an unbudgeted request for more money, a resignation on the critical path — and the board asks the question that always seems to arrive too late: how did we not see this coming?
We did see it. The signals were there the whole time. They simply were not in the report, because the report had been built to measure what had already happened.
The Two Clocks
Every measure a programme keeps runs on one of two clocks. A lagging indicator records an outcome that has already occurred: a milestone delivered, a cost variance booked, a defect found in system test, a benefit realised or missed. It is history, and its great virtue is that it is certain — the event is in the past and cannot be argued with. A leading indicator faces the other way. It is a signal that precedes the outcome it foretells — the rate at which requirements are still changing, the length of time decisions sit unmade, the age of the oldest open risk, the turnover of people on the critical path, the volatility of the team’s own estimates from one month to the next. Its great weakness is the mirror image of the lagging indicator’s virtue: it is uncertain. It describes something that has not happened yet and therefore might not.
Put plainly, lagging indicators tell you where you have been; leading indicators tell you where you are going, and charge you the price of not being sure.
| Lagging indicator | Leading indicator |
|---|---|
| Milestones completed this period | Decisions requested but not yet made, and how long they have waited |
| Cost and schedule variance booked to date | Volatility in the team’s own estimates from one month to the next |
| Defects found in system test | Rate of requirement change after a baseline is set |
| Benefits realised against the case | Turnover of key people on the critical path |
Why “Add Some Leading Indicators” Fails
The textbook answer to all this is by now well rehearsed. Balance your scorecard. Do not steer by the rear-view mirror. Complement your lagging measures with leading ones so that you are guiding the programme rather than merely recording its history. It is sound advice, and almost no one disputes it. And yet the organisations that set out to follow it mostly end up no better warned than before. The reason is worth dwelling on, because it is not the reason the textbooks give.
The first failure is that we manufacture lagging indicators with an earlier timestamp and call them leading. “Milestones achieved this month” does not become a leading indicator merely because the month falls early in the plan; it is still a record of something that has already happened. Moving the clock forward does not change which direction it faces.
The second failure is that we choose the measurable over the predictive. Faced with the question “what shall we track?”, the instinct is to reach for what is easy to count — headcount deployed, percentage complete, documents signed off — rather than what actually foretells trouble. Percentage complete is the great offender here: it is comforting, it is easy to report, and it is very nearly useless, which is why programmes can sit at ninety per cent complete for a third of their duration.
The third failure is the most human of the three. Once a leading indicator carries a consequence — once turning it amber invites a difficult conversation — people begin to manage the indicator rather than the reality beneath it. The signal is smoothed, the definition is quietly renegotiated, the amber is held off until it is undeniable, at which point it has stopped leading anything at all.
The Signal and the Nerve
Here is the lesson experience teaches and the textbooks tend to miss. The scarce ingredient in early warning is not the signal. Any seasoned hand can list the measures that genuinely lead — decision latency, rework rate, the ageing of open risks, the churn in a team’s own estimates. They are not secret and they are not hard to capture. The scarce ingredient is the willingness to act on one of them while the outcome it points to is still uncertain and still deniable.
A signal is only leading if someone is willing to act on it while it can still be denied. That is the whole of the difficulty. A lagging indicator demands nothing of anyone’s nerve, because by the time it turns red the event has already happened and action is merely tidying up after it. A leading indicator always asks the opposite — that you intervene in a programme which still looks well, spend political capital, challenge a colleague’s optimistic report, slow something down or reopen a settled decision, all on the strength of a number the programme’s own sponsor can plausibly wave away as noise. Which is why the distinction between a leading and a lagging indicator is not really a property of the metric at all.
“The distinction between a leading and a lagging indicator is not written in the metric; it is written in what the organisation does the morning the number first moves.”
A Number That Was There All Along
Consider a composite that will be familiar in outline to anyone who has run a large change programme. The programme was substantial — a multi-year business change effort with a heavy systems component, a full-time office, a monthly board. For four consecutive reports it was green. Every milestone was met, if sometimes by visible heroics in the closing fortnight, and the earned-value figures were respectable.
One measure, kept almost as an afterthought by a diligent planner, told a different story: the average age of the decisions the programme had formally asked of its steering group and not yet received. Over two quarters that figure climbed from nine days to thirty-four. The programme was still hitting its milestones, but it was doing so by making local assumptions in place of the decisions it was waiting for, and the stock of unmade decisions was quietly compounding beneath the green surface. That number was visible a full quarter before the first milestone slipped. No one acted on it, because it was not red — there was no threshold, no rule, no consequence attached to it, and every other indicator was reassuringly green.
When the effect finally reached the milestones it did so all at once, as suppressed problems tend to. Three of the assumptions made in lieu of decisions proved wrong; the rework carried past a stage gate; something close to a quarter’s runway had already been lost by the time anyone was permitted to say so out loud. The signal had done its work perfectly. The organisation had simply declined to listen to it while listening would still have cost something.
But Isn’t This Just Earned Value Done Well?
The strongest objection to all of this comes from the earned-value tradition, and it deserves a fair hearing rather than a straw man. Earned value management, properly run, is genuinely more forward-looking than counting milestones. A schedule performance index drifting below one, a cost performance index sliding month on month, an estimate at completion creeping steadily upward — these do extrapolate from the present toward the finish line, and a programme office that watches their trends rather than their monthly snapshots holds a real early-warning instrument in its hands. Is the argument here not simply a case for doing earned value well?
Up to a point, and the point is where it matters. Earned value is among the most useful instruments a programme has, and nothing here is an argument against it. But it measures the physics of the plan, not the health of the assumptions beneath the plan. A programme can post an immaculate schedule performance index while building, with great efficiency, something the business has quietly stopped wanting — or while resting on a decision everyone is avoiding. Earned value tells you whether you are on course to deliver the plan you have; it cannot tell you whether that plan still describes reality, and it cannot tell you whether the organisation has the nerve to act on what the plan is beginning to whisper. The decision-age number in the composite above would not have surfaced in any performance index until it had already become cost and schedule. The leading signal lived precisely in the gap that earned value, by its nature, does not reach.
What To Actually Do About It
None of this argues for a larger dashboard. The instinct, having read this far, is to go and assemble a battery of predictive metrics — and it is the wrong instinct, because a wall of leading indicators that no one has agreed to act upon is merely a more elaborate way of being surprised. The practical lesson is narrower, and harder.
Choose very few genuine leading signals — two or three, no more — and choose them for their predictive power rather than the ease of counting them. Publish them alongside the RAG status, not buried three appendices beneath it, so that the contrast between the reassuring surface and the moving early signal sits in the same glance. And then do the one thing that actually matters and that no measurement can do on your behalf: decide, in advance and in the calm before anything has gone wrong, what the response will be when one of those signals turns — who convenes, what is reopened, what is allowed to slow down — so that acting early becomes a pre-agreed reflex rather than an act of individual courage in the moment it is least available.
Because the honest reframing is this. The question worth asking of a programme is not “what are our leading indicators?” That question is easy, and answering it changes nothing. The question is: what are we actually prepared to do about a warning while it can still be argued away? A programme that can answer the second question does not need many indicators. A programme that cannot will not be rescued by any number of them.