Site search is where a customer goes when navigation has already failed them. Failing twice in a row is what makes it expensive, and internal testing almost never reproduces it because internal testers know the words.
What to put in the public brief
- That it is a set of search tasks on a website
- Desktop or mobile
- That screenshots are required
- Roughly how long
- Whether familiarity with the category disqualifies them
Keep the URL and any query list private.
What to put in the private instructions
Give goals, never queries
"Find something to keep a laptop cool" rather than "search for laptop stand". The words they pick are the finding.
Ask them to record the exact query typed
Character for character, including anything they abandoned halfway.
Ask what they did after a bad result
Refine, go to the menu, use a search engine instead, or leave. The recovery path tells you what failure actually costs.
Run typos as a second pass
Ask them to repeat one successful search with a plausible misspelling and report what happens. Zero results for a doubled letter is a common and quietly costly defect.
Proof worth requiring
Ask for
The exact queries typed, screenshots of the results, whether they found it, and what they did next when they did not.
Do not ask for
A judgement on result relevance ranking. Ordinary users cannot assess ranking quality in the abstract, and the useful signal is behavioural: what they clicked, or that they did not.
When this task type is the wrong tool
If people never use search, find out why before improving it — a findability test tells you whether the menu is doing its job.
If search works and the product genuinely is not there, that is a catalogue problem wearing a search problem's clothes.