1. Define the job before choosing a channel
Start with a sentence: ‘For [specific user], when [trigger] happens, our product helps them [complete a job] without [current difficulty].’ If the sentence could describe almost any business, narrow it.
For an illustrative research product, ‘investors’ is too broad. ‘An individual investor maintaining a monthly strategy in spreadsheets and struggling to review changes consistently’ gives you a task, a workaround, and language to search for. This example describes a hypothesis, not a verified buyer.
Write three exclusion rules too. Someone asking for an unsupported feature, a different market, or a completely free service may be a useful interview but a poor initial commercial prospect.
2. Search for the workaround and the trigger
Search where your intended users discuss their work: public community threads, product-evaluation pages, professional articles, and relevant issue trackers. The best source depends on the audience. Do not assume every buyer talks on the same platform.
Combine the job with language that suggests a decision or recurring frustration. Inspect the original page and its date; a search snippet alone can hide the context. Check whether a question was already resolved before responding.
- "[job]" "spreadsheet" "manual"
- "[category]" "looking for" "recommend"
- "[current tool]" "switching"
- "[task]" "takes hours"
- "[category]" "alternative" "need"
3. Build a small, reviewable shortlist
For each candidate, save the source URL, date, exact problem, current alternative, and your reason for considering them. Separate what the person said from what you inferred. ‘Asked for recommendations’ is observed. ‘Has a budget’ requires its own evidence.
Review five to ten strong candidates before widening the search. Ask whether the problem is real, your product fits, the timing is current, and a legitimate contact path exists. Rejecting a weak match is progress: it protects your time and keeps the experiment interpretable.
A weighted score can order the queue, but it cannot create missing evidence. A recent source with a clear problem is usually more actionable than a company profile with no relevant trigger.
4. Ask a source-specific question
Offer something that helps with the problem already being discussed. Respect the community’s rules and the person’s boundaries. If a public response is appropriate, answer the question before mentioning your product. If there is a professional contact route, keep the message short and relevant.
Illustrative opener: ‘You mentioned rebuilding the report every month. Is collecting the inputs or comparing changes taking most of the time? I am working on that workflow and can share a small example if useful.’ This tests context without pretending you know their budget or intent.
Review the exact copy and destination before sending. Keep a record of what was sent and when. A lack of response does not justify unlimited follow-ups; set a modest stop rule before starting.
5. Define what a useful trial looks like
A signup is an entry point. Activation is the first useful result. For a research product it might be a completed analysis; for a learning product it might be correcting an error and succeeding on a new question. Choose the event that shows someone experienced the product’s value.
Invite the person to one specific action, then observe where they get stuck. Ask about their existing process and what would make the new approach worth returning to. Avoid leading questions that assume the person likes your solution.
| Stage | Record | Do not substitute |
|---|---|---|
| Conversation | A substantive reply about the problem | Message sent |
| Activation | A completed useful product action | Page opened |
| Retention | A meaningful return to the workflow | An email click |
| Customer | A real paid transaction or recorded agreement | Estimated prospect value |
6. Make one decision at the end of the experiment
Choose a review date and inspect the full sequence: who you approached, why they fit, what you asked, what happened, and where the next step failed. Keep the denominator. Three replies from five carefully selected people means something different from three replies from five hundred generic messages.
If the audience recognizes the problem but does not try the product, change the offer or reduce the effort required. If people activate but do not return, investigate whether the product solves a recurring job. If the problem is absent, revise the audience. Change one major variable at a time.
Use the acquisition-plan worksheet to keep the experiment small and explicit. DemandBolt can carry the same brief into a project, then keep evidence, approvals, and outcomes together.