Where Projects Actually Fail.

Written and narrated by Serguei Poppeleer.

Ask where a project failed and people point to the gate where the problem surfaced. But that is where it was seen — not where it happened. It happened weeks earlier, between gates, while everyone was still looking at the last green light.

A project does not fail at the checkpoint. It fails in the space between checkpoints — and the longer that space, the more room there is to fail in it.

A checkpoint tells you the state of the project at the moment you looked. It is a photograph. It says nothing about the next four weeks, when a lead time doubles, a rule changes, or an assumption your business case rested on stops being true. By the next gate, the damage is done and the recovery is expensive.

Take the real shape it makes on the ground: the steel is confirmed in March, the programme is built on that date, and the April gate shows green. In May the supplier’s government tightens an export rule and the order sits stuck at the border. Nothing was wrong at the gate; everything was wrong three weeks after it.

And the context is no longer standing still between your reviews. Supply chains are regionalising, jurisdictions diverging, trade rules moving on political time. The assumption that you can assess once and execute against a stable world — the once-a-year control posture — is quietly becoming the most expensive belief in the business.

The question is no longer only “are we on track against the plan?” It is “is the plan still pointed at the right target?” — and that has to be asked continuously, not once a year.

The belief that creates the gap

The belief is that control is something you confirm at intervals — pass the gate, update the log, file the report, move on until the next checkpoint. Periodic control made sense in a stable world where little changed between reviews.

We do not work in that world anymore. The further a programme reaches across borders and the faster conditions move, the more reality diverges from the plan between gates. Frontier markets do not create this — they just expose it faster. Now the exposure is spreading to markets that used to feel safe.

Periodic control assumes a stable world. We stopped living there, but our review cycles did not notice.


Control as a living system, not a checkpoint

The fix is to move from confirming control at intervals to running it continuously — a standing rhythm that reads the project and the market as they move, and acts before the next gate. The 5Cs are designed for exactly this: mechanisms that run every week, not artefacts you complete every quarter.

Continuous
A standing cadence, not a calendar gate

The control mechanisms run on a weekly rhythm, so drift surfaces in days, not at the next milestone review.

Assumptions
A ledger you actually watch

The assumptions the business case depends on are written down and monitored, so you see the one that just stopped being true.

Triggers
Pre-agreed signals that force a change

Decide in advance which market or supply signals require the plan to be re-pointed, so reaction is a routine, not a crisis.


What to do this quarter

1
Measure your gap

How many weeks pass between the moments you genuinely re-check the plan against reality? That interval is your exposure window.

2
Write the assumption ledger

List the assumptions your case rests on and the conditions that would break each. You cannot watch what you have not named.

3
Set the triggers in advance

Agree now which signals force a re-plan, so the decision is made calmly before the shock, not anxiously after it.

4
Shorten the loop

Replace the quarterly re-assessment with a standing rhythm. The cost of constant attention is now lower than the cost of waiting for the next review.


The objections

“Continuous control sounds like constant firefighting.”

It is the opposite. Firefighting is what happens when you only look at the gates and discover the fire late. A standing rhythm surfaces small issues early, while they are still cheap — so there are fewer fires, not more.

“Our markets are stable enough to plan annually.”

For now. The point is not to predict which shock arrives, but to build a loop that catches whichever one does. The once-a-year control posture only looks fine until the year a single assumption moves between your reviews.

AI Reality Check

Continuous control used to be unaffordable — you cannot put a person on every assumption and lead time, re-checking weekly. That is precisely why we settled for periodic reviews. AI changes the economics: it can watch the inputs continuously and surface the one that just moved against your plan.

But it watches the assumptions you defined and the plan you gave it. It does not decide what matters or what to do about it. It closes the gap between checkpoints by making the watching cheap; the judgement at the trigger stays human.

AI can read the market between your gates. It cannot tell you which way to steer when it moves.


What it prevents

A cross-border EPC programme — €850M+, anonymised

Across multiple jurisdictions and a shifting contracting environment, the dangerous failures were never at the formal gates — they were the conditions that changed between them. When a regulatory change landed on the supply route mid-programme, the standing rhythm surfaced the exposure within days — the next formal gate was still six weeks away. That short loop is what kept the programme defensible when the environment pushed back.

Nothing was caught because a gate arrived. It was caught because the loop was short enough to see the change before it became a claim.

The gate is where you find out.

The space between gates is where it was decided. Close that space, and you stop being surprised by your own project.

Where does your delivery system actually stand?

The Delivery Score measures your governance architecture across the five dimensions of The 5Cs Framework™. Fifteen Yes/No questions. About four minutes. Instant result.

Get my Delivery Score — free → Back to the library →