QIPM Primer · Free

The status that was never true

Three chapters, one exercise, no framework required.

This primer is free, and it is complete. It contains one exercise you can run on your own team this week — no permission, no tool, nothing to adopt. If you read only chapter 3 and do the exercise, you have got the value.

It gives away two of QIPM's five laws. That is deliberate: these two stand on their own.


1The status

The status that was never true

Thursday, late morning. A project manager needs to know whether the payment integration will be ready for the release.

She asks the developer, who says it is done — code merged, tests passing, he moved on two days ago. She asks the tester, who says it has not started — it is not in the test environment, and nothing has been deployed there since Monday. She asks the platform engineer, who says it is blocked, because the configuration change it needs is waiting on a security review nobody has scheduled.

Three answers. Nobody is lying, nobody is confused, nobody is even being imprecise. All three answers are correct.

So she does what the tooling asks of her. She picks one. The board says In Progress, the weekly report says Amber, and the release meeting is told the integration is "on track with a dependency." Every one of those statements is less true than any of the three answers she was given — and each is now the official position.

This happens on every project, every week, everywhere. It is so ordinary that we have stopped noticing it is strange.

What went wrong is not what we think went wrong

The instinctive diagnosis is a discipline problem. The definition of Done is too loose, the board is not maintained, the handover is sloppy. Someone should tighten something.

Every project manager reading this has tried that fix. Most have tried it more than once. It does not work, for a reason that has nothing to do with discipline:

The item does not have a state. It has a distribution over states, and each observer holds a different slice of it.

The developer's slice is real: the code exists and behaves. The tester's slice is real: no evidence has been produced. The engineer's slice is real: a gate is closed. There is no hidden fourth answer that reconciles them, and no meeting that would have surfaced "the truth" — because the thing we call status was never a property of the work. It was a property of the relationship between the work and whoever is looking at it.

A tighter definition of Done does not remove that. It moves the disagreement to a different word.

Two consequences, and they are not philosophical

Done is an event, not a column. An item does not sit in Done. It collapses into Done at a moment in time — when the last observer's slice agrees with everyone else's. A board that treats Done as a place where work rests is recording an opinion and presenting it as a fact.

The ambiguity is structural, so the only useful questions are how much of it there is and how long it lasts. Not "what is the status," which has no answer, but "how confident are we that this collapses by the date we care about" — which does.

That single substitution is the whole of it. A status is a claim about the present that is never quite true. A confidence is a claim about the future that can be checked afterwards.

And checked afterwards is the part that matters. Anyone can say 70%. What makes it worth more than "on track" is keeping score — which is why this framework is not simply a nicer vocabulary for hedging.


2The gap

Why the framework you already have cannot fix this

This is not an argument that your framework is bad. Scrum, Kanban, SAFe and PMBOK are each good at what they were built for. The argument is narrower and, I think, harder to dismiss: all four are silent on the same thing.

Scrum is a framework for flow and feedback — a fixed cadence, protection from mid-sprint churn, four points where reality gets inspected. What it says about an individual item is: it is in the Sprint Backlog, or it meets the Definition of Done. Two states, binary, by design. The three answers from chapter 1 have nowhere to live.

Kanban is a framework for flow under constraint, and the most honest of the four about uncertainty, because it forecasts from measured throughput rather than estimates. But its unit of truth is still the card's position on the board — which is exactly the thing three observers disagree about.

SAFe scales coordination, and solves a real, hard problem reasonably well. What it does with confidence is aggregate it upward — and aggregating a quantity nobody measured accurately at the bottom does not improve it. It makes it look official.

PMBOK is the only one of the four that treats uncertainty as a first-class subject, and compensates with documentation and control. But its instrument is percent-complete against a baseline, and percent-complete is the most confidently false number in project management.

None of these is a failure of design. Each is a rational response to the constraint its authors faced. The point is what falls in the gap between all of them.

Three teams at 90% are not at 90%

Here is the arithmetic that lives in the gap. It takes ten seconds and it changes how you read every dashboard you will ever see again.

Three teams each hold one link of a chain. Each reports, honestly and accurately, 90% confidence in delivering their part.

0.90 × 0.90 × 0.90 = 0.729

The chain is at 73%. Not 90%. Add a fourth link and you are at 66%. A fifth, 59%. By the eighth, a thing every dashboard shows as near-certain is a coin flip with better manners.

LinksEach reportsChain actually is
190%90%
290%81%
390%73%
490%66%
590%59%
690%53%
890%43%

Confidence along a dependency chain multiplies. It does not average, and it certainly does not take the maximum. Yet every status roll-up in common use effectively averages: eight green teams produce a green programme.

I have watched this play out on a real delivery — eleven work items across five teams, every owner reporting in good faith. The mean item confidence was 84%. The delivery was at 44%. Nobody lied. Nobody was even wrong. The entire error lived in the space between the reports, which is precisely the space that no status process measures.

Two refinements worth knowing, because the naive version of this argument is easy to attack.

Not every dependency is a hard block. Some slips stop you dead; some cost you an afternoon. Multiplying everything as though it were a hard block makes the map cry wolf, and a map that cries wolf stops being read. The full framework weights each link by how strongly a slip actually propagates.

Paths share items, so do not multiply the paths together. Two delivery paths that both depend on the same integration are not independent; treating them as independent double-counts the same risk and produces a number so pessimistic nobody believes it. Take the weakest path instead. Being conservative and credible beats being alarming.

That is the first thing QIPM does: it makes the 40-point gap visible before the quarter ends rather than after. The second thing is the next chapter — and it is the one you can do today.


3The exercise

The number almost nobody has measured

Here is a question with a real answer that your organisation almost certainly does not know:

How many hours per week does your team spend being measured?

Not delivering. Not planning. Not even meeting about the work. Specifically: producing, assembling, maintaining and defending information whose purpose is to inform someone outside the doing of the work.

Nearly every organisation tracks its licences and its cloud bill to two decimal places, and has never once put a number on this.

Why this is a cost, not just an annoyance

Measurement is not a window onto the project. It is an intervention in it, and it charges twice.

It consumes. Hours in reporting meetings, preparing for them, maintaining fields for other people's benefit, assembling decks, answering "quick questions" — all subtracted from the work being reported on. Preparation is usually the larger half, and it is almost never counted.

It distorts, which is worse. An item reported as on-track becomes harder to report as off-track — not through dishonesty, but because the act of observation converted an estimate into a commitment. The tighter and more frequent the observation, the more the work optimises for looking measurable rather than being finished. Anyone who has watched a team groom a board the night before a steering committee has seen this happen.

Neither cost appears in any budget. Both are paid out of delivery.

The exercise — 30 minutes, alone

You can do this alone — estimating it yourself is accurate enough, and waiting for a mandate to count something is how counting it gets postponed forever. Nothing here commits anyone to anything: it is a measurement, not a change.

List every observation whose purpose is to inform someone outside the team: recurring reporting meetings and their preparation, fields maintained for others, reports and decks, the standing "when will X be ready?" traffic, timesheets and periodic attestations. For each, estimate hours per occurrence, occurrences per week, and how many people are involved. Then add one column that matters more than it looks: who owns this? — the person who would notice if it stopped. Leave it blank when there isn't one.

hours/week       = h/occurrence × occurrences/week × people
Observation Load = total hours/week ÷ (team size × 40)

Calculator

Observation Load calculator

Pre-filled with a worked example — edit it, or clear it and use your own.
Observation h / occ per week people h / week Owner Remove
total h / week
h / person / week
observation load
unowned observations
Enter your own observations above to see where you stand.

What to do with your number

Do not launch a reform programme. Do three things.

Publish it, without a proposal attached. Just the number, to your team and your manager. It changes the conversation on its own — and a number that arrives with a demand gets argued with rather than absorbed.

Delete one unowned observation. Pick the one nobody will defend. See whether anyone notices. Usually nobody does, and that single result is worth more than any argument you could have made.

Then ask the real question. For each remaining row: is this being asked because someone needs to decide something, or because it is Thursday? Observations that inform a decision are worth their cost. Observations that inform a habit are not.

Two things this exercise is not. It is not an argument that reporting is bad — some of it is essential. And it is not an accusation aimed at managers, who are usually asking because the information they get is untrustworthy. That is the actual root cause, and it is what the rest of the framework addresses.


What's next

What the rest of QIPM does with this

Three chapters in, you have two of the five laws — confidence multiplies along a chain, and observation is not free — plus one number about your own team. That is useful on its own, and you owe nobody anything for it.

The rest answers the question those two laws raise: if status is untrustworthy and measuring it is expensive, what do you actually do on Monday?

You publish confidence instead of status, and then you keep score. One number per item, one date it refers to, and a one-line record when it closes of what you claimed versus what happened. After about twenty items that record tells you something no retrospective ever has — not whether you delivered, but whether your 80% means 80%. In the ledger I keep, aggregate calibration looked flawless while the team was 20 points overconfident on third-party integrations and 15 points underconfident on its own internal APIs. Two real problems with opposite fixes, cancelling each other to almost exactly zero. Both invisible without the table.

You stop promising two things at once. Fix the scope or fix the date; you cannot fix both. Demanding both does not remove the uncertainty — it relocates it somewhere you can no longer see it, which is usually quality, or the team.

There is an obvious objection to all of this, and it is the right one to raise: what do I do when 45% is politically unacceptable? Every project manager has been in that room. The framework's answer is an order of operations rather than a slogan — track record first, then the band, then the choice — because a probability offered by someone with no record is a hedge, while the same figure offered by someone whose 70% has been 70% twenty times is a commitment. The Guide devotes a chapter to that conversation and to the fourteen other objections worth taking seriously.

And you let the machine do the observing. This is the part that is new rather than rearranged. Every agile framework was built around the assumption that measurement is expensive because humans perform it — and that assumption expired recently. A model can read your board, your commits and your CI results and propose each item's confidence without interrupting anyone. So the honest answer to chapter 3 is not "measure less." It is that measurement should stop costing human attention, because attention is now the scarce thing.

That is the framework: five laws, five one-page artifacts, one twenty-minute meeting, three metrics, no new roles and no new tools. It sits on top of whatever you already run. Keep your sprints, keep your board, keep SAFe if you have it.

One last thing, on the physics

QIPM borrows structure from quantum mechanics. It is not quantum mechanics, it requires none, and it does not claim projects are quantum systems.

Some mappings are literal: a state as a probability distribution is statistics, not analogy, and confidence multiplying along a chain is exactly the arithmetic in chapter 2. Some are structural analogies, like the trade-off between scope precision and date precision. And some — superposition above all — are frankly metaphors: chapter 1 describes epistemic uncertainty, different observers holding different information, not quantum superposition.

The Guide prints that table on page three, before making any claim, because a framework about honest measurement should start by being precise about itself.