Your quality system is built to prove you produced what you described. It holds no instrument capable of asking whether the description was worth producing.
An aged-care operator opens a new residence. The scenario is hypothetical; the pattern is not. Every inspection passed, every non-conformance was closed out, the accreditation audit was clean. Eighteen months later the operator is funding capital works to move a nurses' station, widen two corridors and convert an activity room into storage, because the way staff actually move between residents at three in the morning bears no relationship to the circulation the design assumed.
None of that was a quality failure. Every element was built exactly as specified, to standard, and signed off by people doing their jobs correctly. The building conforms. It simply does not work.
This is the failure an assurance system is structurally incapable of catching, and most executives have never been shown where the gap sits. A green quality report is read as evidence that the investment is sound. It is evidence of something far narrower — that the organisation did what it said it would do. Whether what it said was worth doing is a question the quality apparatus was never built to ask. The instrument that is capable of asking it usually appears on the calendar as a progress meeting.
The Strategic Context
The word quality does at least five jobs at once. Teaching material in the field routinely lists them: quality as excellence, quality as a bundle of product attributes, quality as fitness for intended use, quality relative to price, and quality as conformance to specification — and then proceeds as though the last were the only one that could be managed.
That is not carelessness. It is a consequence of what can be audited. Conformance is cheap to verify and produces a defensible paper trail. The other four require judgement, and judgement does not photograph well in a governance pack.
Research on construction projects separates the term more usefully into four distinct problems, each needing a different technique. Winch, Usmani and Edkins, writing in Construction Management and Economics in 1998, distinguish the quality of conception — whether the thing makes sense at all, managed through peer and design review; the quality of specification — fitness for purpose, managed through value management and whole-life costing; the quality of realisation — the client's judgement of how the process was run; and the quality of conformance — whether what was built matches the description, which is where assurance and control live.
Their conclusion belongs above every assurance dashboard in the enterprise: consummate realisation of a specification, right first time, will do nothing to close the gap between what the client expected and what they received, if the specification is not fit for its purpose.
What Leaders Commonly Misread
That assurance and control cover the field. They cover one quarter of it, and the cheapest quarter. The research is blunt about the division of labour: assurance and control can only ensure conformance to standards that already exist; it is the successive design reviews that set the standards for conception and specification. An organisation with excellent assurance and weak reviews has built a reliable machine for producing the wrong answer.
That a longer quality plan is a better one. A course handout on project quality management offers as its model a list of eighteen headings lifted from a quality-management standard — document control, management responsibilities, contract review, design control, purchasing, traceability, process control, inspection and test status, control of non-conforming product, corrective action, records, audits and the rest. [FACT CHECK REQUIRED — the clause structure reproduced in that handout appears to predate the 2000 revision of the standard it names] Read the list looking for the heading that asks whether the specification is worth conforming to. There is not one. Every heading governs the fidelity of execution against a requirement that arrived from somewhere else, unexamined.
To its credit the handout undercuts its own list two paragraphs later — the plan should be only as detailed as it needs to be — and two pages earlier states the premise the whole movement rests on: quality "can not be 'inspected in'". The instruction and the instrument point in opposite directions, and in most organisations the instrument wins.
That "do it right the first time" is a complete instruction. It is half of one. Doing it right the first time has value only once someone has established what it is, and that establishment is a separate act, with a separate owner and a separate failure mode.
Reframing the Issue
The useful move is to stop treating quality as a property of the output and start treating it as a sequence of screens, each of which lets a decision through or sends it back.
An initiative is, from one angle, a flow of information that begins as near-total uncertainty and ends as a complete description of a thing. At intervals it passes through decision points that narrow it. Before each point the options are open and cheap; after it they are settled and expensive — and what happens there sets the standard everything downstream is measured against.
So the strategic question is not how good is our assurance? It is: at which events does this enterprise actually set its standards, who attends them, and what are they entitled to change? In most organisations that has never been asked in those words, and the answer turns out to be that standards are set by whoever drafted the requirement and ratified by a meeting that believed its job was to note progress.
The same researchers add an observation worth sitting with. Quality on a project is not delivered but negotiated — a proposition they attribute to earlier commentators and then sharpen with their own evidence, concluding that design reviews are the forum in which the negotiation happens. If that is right, a review treated as a status update is not a weak control. It is an abandoned negotiation, and the other party writes the terms.
Strategic Analysis
Four qualities, four instruments — and most enterprises have bought one
Set the four problems against the apparatus a typical organisation holds. For conformance: an audit function, a non-conformance register, inspection and test plans, a standard. For specification: at best an occasional value-engineering exercise, run once the budget is already in trouble. For conception: a review that may or may not have the right people in it. For realisation: a satisfaction survey issued after everyone has left.
The imbalance is not irrational — conformance instruments defend the organisation in a dispute, and defensibility is a genuine requirement. But a conformance-heavy portfolio buys legal protection and operational discipline, and almost nothing against building a well-made irrelevance. Different risks; different money.
A review has two ways to fail, and both are scheduling decisions
The research names them precisely, and both sit within an executive's gift.
Timing. Too early and the review cannot function as a screen, because there is not yet enough definition to screen. Too late and the work it should have redirected has been done, leaving only acceptance or write-off. The window is narrow and it is not the same width on every initiative.
Attendance. Convene everyone and the meeting becomes unwieldy — and in a live procurement it can compromise the tender outright. Convene too few and the review cannot see all the options, so it screens against a partial picture and calls the result agreement.
Neither is a quality problem. Both are governance design problems that determine quality outcomes, which is why they are invisible to the function nominally accountable for the result. Two adjacent disciplines stay where they belong: [Related article: Which of Your Dependencies Are Real?] owns the logic that fixes a date, and [Related article: Enough Technical Depth to Test the Answer] owns whether a reviewer can falsify what is put before them. This article owns only what a review is entitled to change.
What a serious review architecture looks like
The same paper documents a large pharmaceutical research campus whose client took this unusually seriously. Reviews came in two kinds — formal ones at defined points, interim ones convened as needed — and the formal points were named for the decision each was making rather than for the calendar: a concept review as the master planner handed the control documents to the design consortium; a design adequacy review marking the point at which the definition was complete; a package scope review two months into detailed design; a pre-issue review a week before each package was released. At each, a review record was completed and the manager of every design team committed personally to what had been decided.
Two features matter more than the names. The reviews sit at information handovers rather than at time intervals — where responsibility passes between parties, which is precisely where a standard is most likely to be silently redefined. And each carries a stated entitlement: the concept review could change the concept; the pre-issue review could not, and was confined to coordination and completeness. A review entitled to change everything is unmanageable. A review entitled to change nothing is theatre. The value lies in saying which is which, in advance.
Decision Framework
The instrument is a review register, and it is short. For every screen between an idea and a finished thing, record five things.
| Field | The question it answers | Failure signal |
|---|---|---|
| Class | Which quality is this review entitled to change — conception, specification, realisation, or only conformance? | Every review in the register is a conformance review |
| Position | Which information handover does it sit at? | Reviews run on monthly cadence rather than at handovers |
| Entitlement | What may it stop, send back or approve? | No review in the register can stop anything |
| Attendance | Who must be present for that entitlement to be real? | The people who will operate the result are not listed |
| Record | Who commits to what was decided, by name? | Minutes note the discussion and name nobody |
Two tests follow, and both can be run on a live initiative this quarter.
The entitlement test. Take a recent decision that turned out badly and trace it to the review that could have caught it. If the review existed but had no authority over that class of thing, the failure is architectural and no volume of assurance would have prevented it. If no such review existed, the enterprise has been setting specifications by default.
The last-cheap-day test. For each screen, ask on what date changing the specification stops being cheap and becomes a variation. If the review sits after that date it is not a screen; it is a briefing. Move it, or accept that the specification is fixed earlier than the governance calendar implies. [Related article: The Front End Owns the Outcome] develops the economics of that boundary; the register's job is only to put the review on the right side of it.
From Strategy to Execution
Immediately. Take one initiative in flight and build its review register from the existing schedule. Most organisations discover two things within an hour: they have more reviews than they thought, and nearly all are conformance reviews wearing other names. Reclassify before adding anything.
Over two or three quarters. Give the conception and specification screens the institutional weight the conformance ones already have — a named chair, a standing membership that includes the people who will operate the result, a documented entitlement, and a record with names against commitments. This costs meeting time and nothing else, which makes it the cheapest governance improvement available to most enterprises.
Longer term. Rebalance the assurance spend. An organisation whose entire quality investment sits downstream of the specification is buying a guarantee it does not need at the price of one it does. That is an operating-model decision, and it belongs with whoever owns the enterprise's assurance function rather than with any individual delivery leader.
One caution on scope. This concerns conformance to a specification. Whether conforming to a method says anything about capability is a different question — [Related article: Why "The Principles Apply to Any Project" Is Only Half True] — as is what a maturity rating is for — [Related article: Who Is Your Maturity Rating For?]. Run the three together and you get a general suspicion of standards, which is neither the argument nor useful.
Signals to Monitor
- Non-conformances trending to zero while rework spend rises. The signature of a conformance-only system: the work matches the description, and the description keeps changing.
- Review minutes recording attendance and progress but no decision. A review that decided nothing screened nothing.
- Value engineering appearing only after a budget breach. Specification quality is being managed reactively, at the point where the options are worst.
- Post-implementation changes funded as capital rather than logged as defects. The organisation has accepted the specification was wrong and stopped counting it.
Questions for the Leadership Team
- For our three largest initiatives, at which specific events is the specification actually set — and who signed the standard everything else is now measured against?
- Which of our reviews is entitled to stop work, and when did one last do so?
- If a review concluded the requirement itself was wrong, what would happen next, and who would pay for it?
- Who represents the people who will operate the result, while the concept is still changeable?
- When something we built performs badly, do we investigate the specification, or only the execution?
Closing Perspective
The uncomfortable part of this argument is that nothing has gone wrong. The assurance function is doing what it was built to do, the auditors are right, the reports are accurate — and the failure is invisible precisely because every visible control is green.
An enterprise gets the quality its instruments can detect. If the only instrument is conformance, it becomes excellent at building what it described, never learns whether the description was worth anything, and keeps paying for that lesson in capital, one reconfiguration at a time.
The remedy is not more assurance. It is to recognise that the reviews already in the calendar are the enterprise's only instrument for governing conception and specification, to say plainly what each is entitled to change, and to put the right people in the room while changing it is still cheap. [Related article: Visibility You Cannot Use] takes up the second half of that — whether the people in the room can read what they are shown. And [Related article: Who Is Entitled to Say It Worked?] asks the question that outlasts them both: who, in the end, is qualified to judge whether the thing was any good.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.