Where suitable data exists, agree with whoever commissioned the work what success should look like before writing starts. Measures worth considering:
- Support demand on the documented topic, counted before and after
- Documentation search behaviour, particularly searches returning no useful result
- Task completion, where the product can measure it
- Page-level helpful or not helpful responses, with the free-text field enabled
Then we report against them.
Not every project can measure all four, and some teams do not have the instrumentation for any of them. Where that is the case, say so at the start rather than at the end.
Be careful what you claim from these. Documentation often has real value that is genuinely hard to attribute: the ticket nobody raised, the evaluation that went well because the docs answered the question first. Requiring every project to demonstrate a measurable change overstates what can honestly be attributed to it. Measure where measurement is meaningful, and be honest about the rest.
