Free tools · Elle Anderson · andersonco.uk

The Discovery Interview Kit

Everything needed to run customer interviews that change decisions instead of confirming them. Built for product people who were told to "do discovery" and handed nothing. Print it and take it into the room.

Before: the prep list

A learning goal, written down. "We need to understand how X currently handles Y." Not "validate our idea"; validation goals produce validation theatre.
Five to eight interviews per round. Patterns appear at five; diminishing returns after ten.
The right people, not the reachable people. Include churned, silent, and nearly-bought, or you build the roadmap for fans.
Pairs where possible. One asks, one notes. The asker who also notes hears half.
No demo in the deck. Show your solution and the interview becomes a politeness contest.

During: the question bank

Opening

Walk me through the last time you did the task. Start from what triggered it.
What did you do just before that? And just after?

Behaviour over opinion

You said you'd love a faster way. What have you actually tried?
Show me how you do it today, screen share or over the shoulder.
When did this last go wrong? What did that cost you, in time or money or grief?

Friction and workaround

What's the annoying part nobody has fixed?
What spreadsheet, sticky note, or group chat is quietly holding this together?

Money and priority

What do you spend on solving this today, including time?
If this vanished tomorrow, what would you honestly do instead?
Who else would need to agree before you changed how this works?

What should I have asked that I didn't?
Who else lives with this problem and sees it differently?

Banned questions: "Would you use a product that…?" and "How much would you pay for…?" People are kind and imaginative; both answers are fiction.

After: the debrief, within 24 hours

Three things they said (quotes, their words)
Two things they did or showed (behaviour beats testimony)
One thing that surprised us
One thing that contradicts what we believed
Follow-up owed, if any

Synthesis rules

Observations before interpretations, labelled. "Four of six keep a manual spreadsheet" is an observation. "They need automation" is a guess.
Every pattern cites two sources. One vivid interview is an anecdote wearing a trench coat.
Hunt counter-evidence deliberately, and report "none found" when true.
Frequency is not importance. One devastating churn story can outweigh nine grumbles; weighting needs someone who was in the room.
End with decisions, not decks: what we now believe, stopped believing, will do differently, and must find out next.

Using AI on the pile? The synthesis prompt pattern I use, with its documented failure modes, is in ai-product-practice on GitHub.

Free to use and share, with attribution.More free tools · How I can help