Is a bug really 100x more expensive to fix later? What the evidence says
The famous “IBM Systems Sciences Institute” chart is cited everywhere. Researchers traced it to old internal course notes. Here’s what you can still say about late bugs.
Short answer: the popular “100 times more expensive in production” figure has no solid source. It’s usually credited to the IBM Systems Sciences Institute, but researchers who traced it found it came from internal IBM training course notes with no supporting data, from projects no later than 1981. A production bug can affect customers before it is reported. Track the actual investigation time and customer impact for your product instead of applying a fixed cost multiplier.
How the claim was traced
- Hillel Wayne questioned whether the cited study exists at all.
- Laurent Bossavit traced the chart to course notes from the IBM Systems Sciences Institute, an internal training program, with no data behind the figures.
- A 1987 textbook presented the numbers as authoritative, and later books and articles repeated them.
What you can say instead
| Instead of | Say |
|---|---|
| “Bugs cost 100x more in production” | “Bugs in production reach real users and cost us their time and trust” |
| “Fixing early is 10x cheaper” | “A bug found with clear details takes less time to fix than one we have to investigate” |
Where the real cost comes from
- Time spent reproducing a vague report.
- People affected before anyone reports it.
- Users who leave instead of reporting.
- Back and forth to get the details.
Those are the costs a good reporting setup reduces: a report with the page, device and screenshot, and a short follow-up for the missing detail, arrives ready to fix.