Cost of Quality

Cost of Quality (CoQ) is the total spend caused by quality, both the money spent achieving it and the money lost failing at it. It exists to settle one argument: quality work is not an extra cost added to development, it is a reallocation of money that gets spent either way.

The four buckets

graph TD
    CoQ[Cost of Quality] --> CoGQ[Cost of good quality]
    CoQ --> CoPQ[Cost of poor quality]
    CoGQ --> PREV[Prevention: stop defects being created]
    CoGQ --> APPR[Appraisal: find defects before release]
    CoPQ --> INT[Internal failure: found before release]
    CoPQ --> EXT[External failure: found by users]

Bucket	Spent on	Examples
Prevention	Making defects unlikely	Requirement reviews, training, standards, better tooling, design work, pairing
Appraisal	Looking for defects	Test design and execution, code review, static analysis, CI compute, audits
Internal failure	Fixing what you found yourself	Rework, re-testing, broken builds, delayed releases, scrapped work
External failure	Fixing what users found	Hotfixes, incident response, support load, credits and penalties, churn, reputation

Prevention and appraisal are chosen. Failure costs are inflicted.

The shape of the trade-off

graph LR
    P0[Little prevention and appraisal] --> F0[Very large failure cost]
    P1[Balanced investment] --> F1[Small failure cost]
    P2[Extreme prevention and appraisal] --> F2[Failure cost near zero but total rises again]

Failure cost falls steeply as prevention and appraisal begin, then flattens. Past the flattening point, each extra check buys less than it costs. That point is not universal: for a pacemaker it sits far to the right, for an internal dashboard far to the left.

Why external failure dominates

An escaped defect costs the fix, plus everything the fix drags with it.

graph LR
    E[Escaped defect] --> S[Support tickets]
    E --> I[Incident response and on-call time]
    E --> H[Hotfix branch, build, release]
    E --> RT[Regression re-testing]
    E --> C[Context switching for the whole team]
    E --> T[Lost trust and possible churn]

The engineering time is usually the smallest of these, which is exactly why estimates made by engineers understate it.

A concrete comparison

One ambiguous requirement about discount stacking:

Where it is caught	Cost
Requirement review	20 minutes of conversation
Design	Half a day of rework on the pricing model
Unit test	An hour, caught by the developer who wrote it
System test	Two days, involving three people and a re-run of the suite
Production	Incident, hotfix, refunds for wrongly charged orders, an audit of affected accounts

The work is identical. Only the amount of built material sitting on top of the mistake changes, and that is what you pay to unwind.

Using CoQ without a finance degree

Full CoQ accounting is rarely worth it in software teams. The useful version is lighter:

• Track escaped defects and roughly what each one consumed, including support and incident time.
• Track rework: time spent on work that was already called done.
• When either rises, move money left in the diagram, into prevention rather than more appraisal, since more appraisal only finds the same defects slightly earlier.
• When failure costs are already low and appraisal keeps growing, you are past the flat part of the curve. Delete tests that never fail and checks nobody reads.

The argument this equips you for: cutting a review to save two days is not a saving, it moves that cost into a bucket where it is larger and arrives later.

Check Your Understanding

#

A team cuts requirement reviews to hit a date, and the release ships on time with three production incidents in the following month. What happened in cost of quality terms?

Total cost of quality fell because prevention spend fell

Prevention cost was moved into external failure cost, which is the most expensive bucket

Correct. The work was not removed, only relocated to a bucket that costs more and arrives later.

Appraisal cost rose while prevention stayed constant

Internal failure cost fell because fewer defects were found before release

#

Failure costs are already near zero, yet the team keeps adding checks and the suite keeps slowing. What does cost of quality suggest?

Keep adding checks, since failure cost can always fall further

The investment is past the point where extra appraisal buys less than it costs, so trim checks that never fail

Correct. Cost of quality has an optimum, and it is not maximum inspection.

Move the spend from prevention into appraisal

Replace appraisal entirely with production monitoring

Quiz Progress

• Answered: 0 / 0 (0%)
• Correct: 0 / 0 (0%) Reset