You cannot improve what you do not measure, and ‘the docs feel fine’ is not a measurement. Documentation is a product; give it the feedback loop of one. The good news is that the signals you need already exist in systems you already run — you just have to look at them, and then act.
Measurement is also how you win the argument for investing in docs at all. ‘We should write more docs’ loses to feature work every time; ‘this doc cut that ticket category by 40%’ wins. Numbers turn documentation from a cost into a demonstrable return.
The signals worth watching
- Searches that return no results — the single most valuable signal you have. It is a literal, unfiltered list of what readers wanted and couldn’t find. Each one is a page waiting to be written.
- Support tickets — tag the ones a doc should have answered. Recurring themes are your highest-priority gaps, ranked for you by real customer pain.
- Page analytics — high traffic with high bounce, or lots of time on a page meant to be quick, hints at confusion worth investigating.
- ‘Was this helpful?’ feedback — cheap to add, and the free-text comments are often the most direct insight you’ll get into where a page falls short.
Pull your docs-search queries that returned nothing from the last week. Most teams have never looked. You will typically find three kinds: things you have but named differently (a synonym/search problem), things you should have but don’t (a content gap), and things that aren’t your product (ignore). The first two are a ready-made, demand-ranked backlog — no guessing required.
Close the loop
Signals are worthless unless they drive change. Turn a recurring ticket into a titled how-to; turn a dead-end search into a new page; rewrite the confusing page the analytics flagged — and then check whether the ticket volume or the failed searches actually dropped. That before-and-after is what proves a doc ‘worked’, and it’s the number you take to your team to justify the next investment. A doc set without a feedback loop is guesswork; with one, it compounds.
Tickets show 30 people a month asking how to rotate a leaked API key. You write ‘Rotate a leaked API key’ as a clear how-to, link it from the errors reference and the security page, and add it to search synonyms for ‘key compromised’. A month later, that ticket category is down two-thirds. You now have both a better doc and a hard number proving documentation ROI — that’s the loop that changes how your org values docs.
In the age of AI
If you offer an AI docs assistant, its logs are a superb measurement instrument: the questions users ask it, and especially the ones it answers badly, map precisely to your content gaps and ambiguities. Feed that back into the doc set and both the human docs and the assistant improve together, in a virtuous loop.
You now have the full toolkit
Plan by reader and task; choose the right type; write each type well; direct AI without trusting it blindly; ship it like code; and measure it like a product. That loop — repeated — is what separates documentation that quietly earns its keep from documentation that quietly costs you customers. You didn’t just learn to write docs; you learned to run documentation as the product asset it actually is.
Answer, then press Check. Explanations appear after.
Choose oneWhich is the most valuable ‘what’s missing’ signal?
ReflectPick one recurring support ticket and describe the doc that would prevent it.
This lesson is part of Technical Writing Pro
Enrol to unlock all 12 lessons — $149.