Imagine a new facility that passes every inspection. Every defect has been closed, the certifier has signed off and the quality records are complete. A year later, the owner is paying to move a workstation, widen an aisle and relocate a door, because the way people actually move and work bears little relationship to what the design assumed. Nothing was built badly. Everything conforms to the drawings. The facility simply does not work as well as it should.
That is not a quality failure in the usual sense, which is exactly why quality systems do not catch it. A clean quality report proves that the work matched its description. It cannot prove that the description was worth building. That question is answered, or left unanswered, much earlier: in how requirements are set, which ones matter most and how design reviews are run.
This article explains the different kinds of quality, why businesses tend to invest heavily in one of them and lightly in the others, how to set priorities among requirements, and how to design reviews that can actually change a specification while change is still cheap.
Four kinds of quality
The word quality does several jobs at once. Research on construction projects by Winch, Usmani and Edkins, published in Construction Management and Economics in 1998, separated it into four distinct problems, each needing a different technique:
| Kind of quality | Question it answers | Typical instrument |
|---|---|---|
| Quality of conception | Does the idea make sense at all? | Peer and design reviews |
| Quality of specification | Is the design fit for its purpose? | Value management, whole-of-life costing, reviews with users |
| Quality of realisation | How well was the process run, from the client’s view? | Feedback during and after delivery |
| Quality of conformance | Does what was delivered match the description? | Inspection, testing, audits, non-conformance records |
The researchers’ central point is that excellent conformance cannot close the gap between what a client expected and what they received if the specification was not fit for purpose. Doing it right the first time only helps once someone has worked out what “it” should be.
Why businesses over-invest in conformance
Conformance is easy to verify and produces a defensible record. Inspection and test plans, audit reports and defect registers protect a business in a dispute and keep execution disciplined. Those are genuine benefits. By contrast, judging whether a concept makes sense or whether a specification is fit for purpose requires judgement, and judgement is harder to document.
So many organisations end up with strong conformance controls and weak controls over conception and specification. They become very reliable at building what they described, without ever checking whether the description was right.
Quality means fit for purpose, not more
Quality is often treated as one corner of a triangle with time and cost, to be traded when constraints bite. That view is too narrow. Quality affects revenue, safety, compliance, customer confidence, maintainability and the productivity of everyone who uses the result.
Quality also does not mean exceeding the specification everywhere. Extra precision, features or finish can add cost, complexity and lead time without improving the outcome. A precision fixture and a disposable packaging insert should not be held to the same tolerances. Quality should be proportional to consequence: high where failure is costly or dangerous, and lighter where it is not.
Many defects are faithful copies of weak instructions
When something performs badly, the investigation usually looks at execution. But many problems are faithful implementations of instructions that were incomplete, contradictory or unmeasurable. Where requirements are unclear, delivery teams make their own assumptions. The result can be technically competent and still wrong for the business. Inspection at the end can find that the work matches the drawings. It cannot recover the decisions that were never made.
Build a hierarchy of requirements
A long list of requirements can look rigorous while hiding a lack of priorities. Classify requirements so everyone knows which ones matter most:
- Non-negotiable: safety, legal and contractual obligations, and core business continuity. These are never traded.
- Value-critical: requirements that decide whether customers or operations get the promised benefit.
- Performance: capacity, reliability, accuracy and service levels, with defined targets and tolerances.
- Preference: convenience and usability features that can be traded if constraints tighten.
Each requirement should be traceable to a reason: customer value, operational need, safety, regulation or strategy. A requirement that cannot be traced to anything deserves challenge. Without a hierarchy, teams may protect minor preferences while compromising what really matters.
Spend on prevention deliberately
The costs of quality fall into three groups. Prevention includes clear requirements, prototypes, supplier qualification and good process design. Appraisal includes reviews, testing and inspection. Failure includes scrap, rework, delay, warranty claims and lost trust, both before and after delivery. No business can eliminate all three. It can choose where to spend. Money spent early on getting the concept and specification right is usually far cheaper than money spent later correcting a well-built mistake.
Design reviews are where specifications are set
If conformance controls check the work against the specification, design reviews are the main instrument for checking the specification itself. Many reviews fail to do that because they are run as progress updates. Two design choices decide whether a review can work.
Timing. Too early, and there is not enough detail to judge. Too late, and the work the review should have redirected is already done, leaving only acceptance or expensive change. Reviews work best at information handovers: points where responsibility passes from one party to another, such as from concept to detailed design, or from design to procurement. Those are the moments when requirements are most likely to be silently reinterpreted.
Attendance. Invite too many people and the review becomes unwieldy. Invite too few and it cannot see the options. The people who will operate, maintain and use the result should be present while the concept can still change.
Two further elements make reviews effective:
- Entitlement: what the review is allowed to change. A concept review should be able to change the concept. A final pre-issue review may be limited to coordination and completeness. A review entitled to change nothing is theatre. A review entitled to change everything at every stage is unmanageable.
- Record: decisions recorded with named people committing to them, not minutes that note the discussion and name nobody.
A review register
A review register lists every review between an idea and a finished result, with five fields:
| Field | Question | Warning sign |
|---|---|---|
| Class | Which kind of quality can this review change: conception, specification or only conformance? | Every review is a conformance review |
| Position | Which information handover does it sit at? | Reviews held on a calendar cycle rather than at handovers |
| Entitlement | What can it stop, send back or approve? | No review can stop anything |
| Attendance | Who must be present for that entitlement to be real? | Users and operators are not listed |
| Record | Who commits to what was decided? | Minutes record discussion but no decisions |
Two tests help check the register:
- The entitlement test: take a past decision that turned out badly and trace it to the review that could have caught it. If that review existed but had no authority over the issue, the problem is the review design, and more inspection would not have helped.
- The last-cheap-day test: for each review, ask the date after which changing the specification stops being cheap and becomes a costly variation. A review held after that date is a briefing, not a control.
A quality plan that connects
A useful quality plan links five things: the outcome and what drives its quality; the requirements and standards that apply; the measures, tolerances and evidence of acceptance; who is responsible for delivery, review, assurance and acceptance; and how non-conformances and changes will be handled. These need to appear in schedules, budgets, supplier contracts and reporting. If reviews and tests are not in the plan with time and money allocated, they are not really planned. The article on inspection and test plans for supplier work covers the conformance end of this chain.
A worked example
This is an illustration. A small manufacturer is moving to a larger building and fitting it out for production. A builder will carry out the works to drawings prepared by a designer. In the first version of events, the owner approves a layout drawn from the old factory, the builder builds exactly to the drawings, every inspection passes and the move happens on time.
Within months, problems appear. The forklift route crosses the main pedestrian path at the dispatch door. The wash bay drains toward the wrong side, so water runs under a workbench. The new delivery supplier’s trucks are taller than the roller door. Power outlets sit where benches cannot go. Fixing them costs about $68,000 and weeks of disruption. None of these was a conformance failure.
In a better version, the owner sets up a simple review register before the design is finalised:
- Concept review, at the handover from layout concept to detailed design. Attended by the production lead, a forklift driver, the dispatch coordinator and a maintenance technician. Entitled to change the layout. The team marks the proposed layout with tape on the floor of the existing building and walks through a normal day, including deliveries and cleaning.
- Specification review, before the builder is asked to quote. Entitled to change door sizes, services locations and drainage. Delivery suppliers are asked for their vehicle dimensions.
- Coordination review, before construction. Limited to checking that drawings are complete and consistent.
Each review records its decisions and the name of the person responsible for each. The reviews take about three half-days of staff time and a few hundred dollars of tape and printing, plus an hour of the designer’s time per review. The layout problems are found and fixed on paper.
How this applies to a small Australian business
Small businesses often rely on builders, designers, certifiers and suppliers for quality. Those parties play essential roles, but they mostly check conformance: that the work meets the drawings, codes and standards. Fitness for your purpose is yours to own. Practical steps:
- Write down what the result must achieve before design starts, in terms of how it will be used.
- Classify requirements as non-negotiable, value-critical, performance or preference.
- Hold reviews at handovers, with users and operators present while change is cheap.
- Say what each review can change, and record decisions with names.
- Mock up or walk through layouts and processes before they are built.
- Confirm regulatory requirements with your certifier or the relevant authority, recognising that compliance is a minimum, not a guarantee of fitness for purpose.
The articles on when your specification becomes the benchmark and choosing your decision measures before the design cover related ideas.
Signals worth watching
- Non-conformances falling while rework and changes after delivery rise.
- Review meetings that record progress but no decisions.
- Value engineering appearing only after a budget overrun.
- Changes after handover funded as new projects rather than recorded as specification problems.
- Requirements still changing after major design commitments.
- Acceptance tests written late in delivery.
- Users discovering constraints the designers did not consider.
Common mistakes
- Treating a clean quality report as proof the investment is sound.
- Investing only in conformance controls.
- Running design reviews as status updates.
- Holding reviews too late to change the specification cheaply.
- Leaving out the people who will use the result.
- Treating all requirements as equal.
- Investigating execution only when something performs badly.
Frequently asked questions
Does this mean inspection and testing are less important? No. Conformance controls remain essential for safety, compliance and disputes. The point is that they need to be paired with controls over conception and specification.
How many reviews does a small project need? Often two or three: one to test the concept with users, one to test the specification before quoting or purchase, and one to check coordination before work starts.
What if a review finds the requirement itself was wrong late in a project? Treat it as useful information, not an embarrassment. Estimate the cost of changing now against the cost of living with the problem, decide openly and record the lesson so the next project’s reviews catch it earlier.
Who should chair the reviews? Someone with authority to make the changes the review is entitled to make, usually the owner or the person accountable for the outcome.
Questions to ask
- At which events is our specification actually set, and who approves it?
- Which of our reviews can stop or send back work, and when did one last do so?
- Are the people who will use the result present while the concept can still change?
- Which requirements are non-negotiable, and which are preferences?
- When something performs badly, do we examine the specification as well as the execution?
- Is our quality spending concentrated after the specification, where it can only check conformance?
Bringing it together
A business gets the quality its instruments can detect. If its only instruments check conformance, it becomes excellent at building what it described without learning whether the description was right. Separate the four kinds of quality, set a hierarchy of requirements, spend on prevention, and design reviews that sit at information handovers, include users, have a defined entitlement to change things and record decisions by name. Put the right people in the room while changing the specification is still cheap.
Source: KEVOS notes, drawing on G. Winch, A. Usmani and A. Edkins, “Towards total project quality: a gap analysis approach”, Construction Management and Economics (1998). Examples and figures in this article are illustrations.
