9 June 2026
No IT project announces its own derailment. It does not turn red on a dashboard: it slips, meeting after meeting, until the day nobody dares to give a go-live date any more. The good news is that the warning signs are remarkably consistent from one project to the next. Here they are, from the quietest to the loudest.
"We are at 90%" for two months now
The great classic. Progress that plateaus at 80 or 90% for weeks measures nothing: it hides the remaining 10%, which almost always contains the hard subjects, namely interfaces, data migration and edge cases. Honest progress is measured in testable deliverables, not declared percentages.
Committees that slip, empty out or go quiet
A steering committee postponed once is life. Postponed three times, it is a symptom. Same thing when decision-makers send substitutes, or when meetings end without a single recorded decision. A healthy project produces decisions. A struggling project produces status updates.
The same questions coming back around
If the data migration scope has already been "settled" three times, it never was: nothing is written down, or nobody owns the call. The simplest test is to ask for the decision register, dated and signed. If it does not exist, you already know the state of the project.
Acceptance testing postponed "to save time"
When the schedule tightens, the first adjustment variable is always the same: testing. It is exactly the wrong move. Reduced testing does not remove defects. It relocates them to production, where they cost ten times more and are visible to everyone.
The supplier's changing faces
The project manager is replaced, then the lead developer, and each time everything has to be re-explained. In theory, turnover at your supplier is not your problem. In practice, your budget pays for every ramp-up.
Budget consumed, no remaining-work estimate
Many projects can say how much they have spent. Very few can say what remains to be done, costed and dated. Yet that is the only question that matters: "remaining work" is the number one steering indicator. If it does not exist, steering does not exist.
What to do when several boxes are ticked
Three reflexes, in order. First, a factual freeze-frame: what is delivered, tested, in defect; the decisions made and missing; the spend and the estimated remaining work. No hunt for culprits, because it freezes everyone and contributes nothing. Second, a re-scoping: in the current state, what can be delivered, by when, sacrificing what? Third, a recovery frame: named responsibilities, written decisions, acceptance testing prepared from now on.
An outside eye helps enormously at that point. Not to judge, but because it can ask the questions nobody inside dares to ask any more. The earlier it steps in, the more options remain open.