Topic 4

What to fix first

Once failures are counted, the obvious move is to fix the most common one. Often that's right, but not always: some failures are rare and harmful, others common and harmless. Weigh how often against how much harm, and both against what the fix costs.

Predict first

Missed retrieval happens 7 times in 20 and makes customers ask again. A wrong fact happens 5 times, and customers act on it. Which do you fix first?

The wrong facts: fewer, but each one does more harmMissed retrieval: it's more commonBoth at onceNeither: 20 failures is too few

What to fix first?

Illustrative counts · severity 1 (annoying) to 3 (harmful)

Fix first

Missed retrieval

Failures

#	Failure	Count	Severity	Impact
1	Retrieval missed the right pageNo answer, or a vague one: the customer asks again.	7	2	14
2	Stated a wrong factThe customer acts on it: wrong price, wrong step.	5	3	15
3	Answered a different questionUnhelpful; the customer asks again.	3	2	6
4	Refused an in-scope questionUnhelpful; looks broken.	2	2	4
5	Too long to read on a phoneAnnoying, but the answer is there.	2	1	2
6	Tone offRude or odd, but the answer is right.	1	1	1

Impact is count × severity. Cost to fix matters too: a cheap fix to a common failure can come before an expensive one to a rare failure.

Ranked by count, missed retrieval is first. Weighted by severity, wrong facts edge ahead: 15 to 14. The scores are close, so cost to fix could easily decide. If the wrong facts come from out-of-date help pages, updating the docs may be quicker than improving retrieval.

Severity is a judgement, so write it down and agree on it, just like a rubric. Lesson 1's link to business numbers helps here. A wrong price on a billing question costs money; a long reply costs patience.

Quick check

A failure is rare but, when it happens, sends a customer's data to the wrong person. Where does it go in the queue?

Bottom: it's rareAt or near the top: severity that high outweighs frequencyWherever the count puts itInto the “other” pile