Defining the Quality Standard

Why the standard comes first#

Most teams measure quality against an opinion. A reviewer "knows good when they see it," which means quality is whatever that reviewer happens to notice on that day. Scores swing with who is reviewing; "quality" becomes a feeling. You cannot improve, audit, or hold a number that isn’t stable — and it can’t be stable until there is a written standard underneath it.

The standard is the foundational artefact of Business Excellence. Measurement scores against it. Audits sample against it. Governance acts on movement in it. Get the standard wrong and everything built on top inherits the flaw.

What a standard must contain#

A usable standard is specific, binary where possible, and tied to what actually matters to the client and the risk. It has four parts.

1. Scope — what it covers#

A standard is for a defined kind of work: a campaign delivery, an approval, a report, a support resolution. Don’t try to write one standard for everything; write one per class of work that matters, in the language of that work.

2. Criteria — the specific conditions of "good"#

The heart of the standard: the concrete conditions a piece of work must satisfy to pass. Each criterion should be stated so it can be judged true or false, not "high quality" or "professional." "Delivered against the agreed line-item specification" beats "delivered well."

3. Severity — not all defects are equal#

A missing legal disclaimer is not the same as a formatting slip. The standard should distinguish the failures that breach it outright (a fail) from minor issues worth recording but not failing on. This stops trivial and serious defects being averaged into a meaningless number.

4. Evidence — how a verdict is reached#

State what a reviewer looks at to judge each criterion, and how the result is recorded. This is what makes coverage provable and scores comparable across reviewers, teams and time.

"High quality" is an opinion. "Meets every criterion in the standard, verified against the evidence" is a fact. Operations runs on facts.

How to write one (without boiling the ocean)#

You do not need a hundred-page manual. You need a tight, agreed definition you can actually apply.

1. Start from failure, not perfection. List the ways work actually goes wrong — the defects clients have caught, the rework you keep doing. Your criteria are mostly the inverse of those.
2. Draft the criteria as pass/fail statements. Force ambiguity out. If a criterion can’t be judged true or false, sharpen it until it can.
3. Assign severity. Mark which breaches are a fail and which are recorded-but-not-failing.
4. Calibrate with real work. Have two reviewers independently score the same sample against the draft. Where they disagree, the standard — not the reviewers — gets fixed. Repeat until they converge.
5. Version it. The standard will evolve as the work and the risks change. Date it, and treat changes deliberately so scores stay comparable.

What it unlocks#

Once the standard exists and reviewers agree on it, the rest of Business Excellence becomes possible:

• Honest measurement — a quality score that means the same thing every time it’s taken.
• Defensible audits — sampling that proves coverage rather than asserting it.
• Root-cause work — defects traced to the criteria they breach, and from there to the process flaw that caused them.
• Governance — a number leadership can steer by, because it trusts what it represents.

This is why the standard is step one. It is unglamorous — no dashboard, no tooling, just careful definition and patient calibration — and it is the step that determines whether everything built on top is measuring reality or measuring noise. Skip it, and you get a quality programme that produces confident numbers about nothing. Get it right, and quality stops being an opinion you defend and becomes a fact you manage.

A worked example#

Abstractions are easy to nod along to and hard to apply. So walk through one, in the generic — a recurring deliverable a team produces and a client signs off on. Call it a client report. The team makes dozens a week. Clients complain about them unpredictably. Nobody can say whether report quality is improving or getting worse, because nobody can say what a good report is beyond "the kind that doesn't draw a complaint."

Start where the method says to start — from failure. Pull the last quarter's complaints and the last quarter's rework. The list comes back concrete: figures that didn't reconcile with the source data; a missing section the client had asked for; commentary that contradicted the numbers above it; the wrong reporting period in the header; a recommendation with no supporting evidence. That is not an abstract wish for excellence. It is a defect catalogue, and the criteria are its inverse.

Now draft each as a statement a reviewer can mark true or false:

• Every figure in the report reconciles to the system of record.
• Every section the client's brief requires is present.
• The narrative commentary is consistent with the figures it describes.
• The reporting period is stated and correct on every page.
• Every recommendation cites the evidence it rests on.

Notice what these are not. None of them says "well written" or "insightful" or "professional." Each one can be checked by a second person who reaches the same answer as the first. That is the test, applied at the level of the individual criterion.

Then assign severity. A figure that doesn't reconcile is a fail — it is the kind of defect that destroys client trust and carries real consequence. A reporting period missing from page four when it is correct on the other pages is worth recording, but it is not the same category of failure, and averaging the two into a single percentage would hide the one that matters behind the one that doesn't. Severity keeps them apart.

Finally, state the evidence. For reconciliation, the reviewer opens the source system and checks the figures against it; the result is recorded against the criterion. For required sections, the reviewer reads the brief and ticks each off. The point of writing this down is not bureaucracy — it is that the next reviewer does exactly the same thing, which is the only reason two reviewers can be expected to agree.

How to know it is working#

A standard earns its place by what it changes, and you can watch for the signs.

Reviewers stop arguing about verdicts. The clearest signal. When two people score the same work and disagree, the conversation used to be about taste — whose judgement was better. With a standard, the conversation is about the standard: which criterion is ambiguous, and how to sharpen it. The disagreement becomes useful, because it points at a specific defect in the definition rather than a clash of opinions that cannot be resolved.

The score becomes stable across reviewers. Hand the same sample to three different people and the verdicts converge. When that happens, the number means the same thing regardless of who took the measurement — which is the precondition for trusting any trend you draw from it.

Defects acquire addresses. A failure is no longer "a quality problem." It is a breach of a named criterion, which traces to a step in the process, which traces to a cause you can remove. The standard turns a vague dissatisfaction into a coordinate on a map.

Disputes get shorter. When a client challenges quality, you are not defending a feeling. You are pointing at a verdict against a written criterion, supported by the evidence the reviewer recorded. The conversation moves from "we think it was fine" to "here is what we checked, and here is the result."

If none of these things is happening — if reviewers still argue, scores still swing, defects are still described in adjectives — the standard is not yet real. It exists as a document but not as an instrument. Go back to calibration.

A standard that lives in a document and not in the reviewers' verdicts is decoration. The test of whether it exists is whether two people use it and agree.

Common failure modes#

Most quality standards fail in one of a few predictable ways. Knowing them in advance is the cheapest way to avoid them.

The aspirational standard. It reads beautifully and scores nothing. "Reports should be insightful, clear, and client-focused." Every word is unobjectionable and not one of them can be marked true or false. This is the most common failure by a wide margin, because aspirational language feels like a standard while asking nothing of anyone. The fix is the pass/fail discipline: if a criterion cannot be judged true or false, it is not yet a criterion.

The hundred-page standard. The opposite failure. In an effort to be complete, the team writes a manual so exhaustive that no reviewer reads it and no two apply it the same way. A standard nobody can hold in their head is a standard nobody uses. Tightness is a feature; the goal is the smallest definition that reliably separates good work from bad.

The uncalibrated standard. Written once, circulated, never tested with two reviewers on the same sample. It looks finished and has never been proven to produce agreement. Drafting is the easy half; a standard that has not survived calibration is a hypothesis, not a standard.

The frozen standard. Written, calibrated, and then never touched again while the work and the risks move on underneath it. Criteria that mattered two years ago no longer match what clients care about now, and the score slowly decouples from reality. The fix is versioning treated as routine — and changed deliberately, so a shift in the number reflects a shift in the work, not a quiet edit to the ruler.

Severity collapse. Every defect treated as equal, so a missing disclaimer and a typo land with the same weight and average into a number that means nothing. Without severity, the score cannot tell leadership whether it is looking at a crisis or a rounding error.

Adapting the standard to your context#

The four parts — scope, criteria, severity, evidence — hold across domains. What changes is how you fill them, and the method is robust to that variation.

In a regulated or safety-critical setting, severity does more work and the bar for a fail sits lower. Some criteria are not "recorded but not failing" under any circumstances; a single breach is a fail regardless of how the rest of the work scored. The standard should make those non-negotiable criteria explicit and separate from the gradeable ones.

In creative or judgement-heavy work, the instinct is to declare the work unmeasurable. Resist it. Even where the core is genuinely subjective, a surprising amount is not: the brief was answered or it was not; the factual claims are accurate or they are not; the legal and brand constraints were respected or they were not. Define the standard around the parts that can be judged objectively, and be honest that the standard governs those and leaves the subjective core to expert judgement. A partial standard applied honestly beats a total standard that pretends the unmeasurable is measurable.

In high-volume, low-margin operations, the constraint is the cost of applying the standard. Here the criteria must be fast to check — favour the few that catch the most consequential failures over a long tail that each catch a little. The standard is shaped by the economics of using it, not only by the ideal of completeness.

Across all three, the discipline is the same: define the work, write criteria you can score, separate the serious from the trivial, and state how a verdict is reached. The contents change with the domain. The structure does not.

How to put this to work#

If you take one thing from this, take the sequence — because the order is the method.

Pick the one class of work where quality matters most and is least defined. Not all of it. One. The deliverable that draws the most complaints, or carries the most risk, or that nobody can currently say anything reliable about. A standard for one thing that actually works is worth more than a standard for everything that works for nothing.

Pull the real failures for that work — the complaints, the rework, the defects clients have caught. Turn each into a criterion stated as a true-or-false claim. Mark which breaches are a fail and which are recorded but survivable. Write down, for each criterion, what a reviewer looks at to reach a verdict.

Then calibrate, because this is the step that separates a standard from a wish. Take a real sample. Have two reviewers score it independently against the draft. Where they disagree, fix the standard — never the reviewers. Repeat until they reliably reach the same verdict. Two or three rounds is normal. The drafting will take an afternoon; the calibration is the work, and it is the work that makes the standard real.

When two people can score the same work and agree, you have something you did not have before: a definition of quality that is a fact rather than an opinion. Measurement can now stand on it. Audits can sample against it. Governance can steer by it. None of that was possible while quality was a feeling, and all of it becomes possible the moment quality becomes a verdict two people share.

That is the whole return on this step. It is the least glamorous artefact in Business Excellence and the one that everything else depends on. Build it first, build it tight, and calibrate it until it holds.