Each department sees a slice of a failure. Production sees scrap, customer service sees the complaint, accounting sees the credit memo, shipping sees the rush freight. Nobody sees the total. Cost of poor quality (COPQ) adds the slices together.
Where it comes from
Armand Feigenbaum set out the four cost-of-quality categories in a 1956 Harvard Business Review article, and Joseph Juran made the cost of failure a management topic in the same decade. The four categories are still the standard model, and ASQ teaches it:
| Category | Examples |
|---|---|
| Prevention | Training, design reviews, error-proofing |
| Appraisal | Inspection, checks, audits |
| Internal failure | Scrap, rework, re-running a report |
| External failure | Returns, warranty, credits, complaint handling, lost customers |
By convention, the cost of poor quality is the two failure categories. Prevention and appraisal are the cost of good quality. The argument for prevention is that it's cheaper than failure.
How a bot helps
For one failure type, such as wrong invoices, a bot traces every place it costs money: the credit memo, the customer service time, the re-billing, the delay in cash, the occasional lost account. It pulls counts from records and multiplies by a cost per occurrence. Each cost per occurrence is labeled: measured (from records), estimated (by the person named), or unknown.
On the bots' own work
A bot's failures cost too: tokens for retries, a person's time correcting output, a rework cycle through another bot. Counting it is how you know whether a better check is worth adding.
When to use it
When deciding which problem to fix first, or making the case for prevention. Medium. Feeds prioritization and the Pareto chart.
Where bots differ
The easy mistake is double counting: the same late payment counted as a cash cost and a lost-revenue cost. Another is inventing a cost per hour. A bot should trace each cost to one category once and leave estimates to a named person.
Sources: ASQ: cost of quality.
Back to Engineering tools for bots.
Short outline
Cost of poor quality: short outline
What it's for
Cost of poor quality adds up what one failure type costs in every department (scrap, rework, credits, rush freight, complaint time, delayed cash), each cost counted once and labeled measured, estimated or unknown.
Columns
| Column | What goes in it |
|---|---|
| Failure type | The failure being costed |
| Department | Where the cost lands |
| Cost element | Scrap, rework, credit, freight, handling time, delayed cash and so on |
| Category | Internal failure, external failure, appraisal or prevention |
| Count | Occurrences in the period, with source |
| Cost per occurrence | With its basis: measured, estimated by a named person, or unknown |
| Total | Count times cost per occurrence |
| Double-count check | Which other element this might overlap, and how it was resolved |
Steps, each with its "done when"
- Pick one failure type and the period. Done when the failure is defined and the period set.
- Trace where it costs, department by department. Done when every department that touches the failure has been asked or its records checked, and each cost element is listed.
- Count and price each element. Done when each element has a count with a source and a cost per occurrence with its basis.
- Remove double counts. Done when each overlap is resolved and noted, and no cost appears twice.
- Total and compare with prevention. Done when the total is shown with its unknowns, alongside what prevention would cost.
Common mistakes
- Counting the same money twice, such as delayed cash as both a cost and lost revenue.
- Made-up cost per hour or per occurrence.
- Costing only one department's slice.
- A total presented without its unknowns.