A week of discovery

Discovery that never decides is performance. A useful week ships no production feature and still de-risks a quarter of work. The cadence is the product trio — PM, designer, engineer — in weekly contact with customers, then a decision with evidence.

Protect the habit like an on-call rotation. Recurring calendar hold. A small panel of customers who said they would talk again. When a launch crunch hits, shrink the session to 20 minutes — do not cancel for three months.

Walking through one week

Outcome: raise week-1 invoice send rate from 40% to 55% among new freelancers.

Monday (30 min). Trio reviews last week's support tags and activation funnel. Three opportunity candidates: unsure what to charge; awkward to request payment details; onboarding feels like paperwork. They pick payment details because last month's interviews showed strong emotion there.

Tuesday (interview). Two users who activated but never sent an invoice. Both stalled when asking clients for card details over email; one used a personal Venmo link and felt unprofessional. Opportunity confirmed in their words.

Wednesday (tree + assumptions). Solutions: magic-link client form; Stripe-hosted checkout only; "copy payment link" with templates. Riskiest assumption for magic-link: clients will complete the form on mobile without support calls. Second: freelancers will trust sending a link instead of attaching a PDF.

Thursday (test). Wizard-of-Oz for five users — engineer hosts a simple form; when a client submits, Slack notifies the freelancer. Kill criterion: fewer than 3/5 freelancers send a link within a week, or more than one client support email per submission.

Friday (decide). 4/5 send links; zero client support tickets; one freelancer asks for branding. Promote magic-link to a thin MVP; park AI pricing under a different opportunity. Update the OST and tell stakeholders with numbers.

That week de-risked a quarter of work. For stakeholders, keep a short log: assumption, test, result, decision. That log is more persuasive than velocity charts when someone asks why engineering is "not building."

Link to delivery: survivors become an MVP that still targets the riskiest remaining assumption. Prioritization then compares opportunities by impact and confidence.

Common mistakes

• Trio of one. Only the PM talks to customers; engineering hears requirements theater.
• Interview theater. Weekly calls that are demos, not research. See user research.
• Infinite discovery. Never shipping a thin slice after assumptions are clear enough to learn in production.
• Big-bang research that pauses delivery for six weeks, then resumes output mode for six months.

Check yourself

Friday's test is 4/5 sent, zero support tickets. What do you do with AI pricing this week?

Park it. It is a different opportunity. Promote magic-link to a thin MVP.

Launch week is on fire. Cancel discovery for a month?

No. Shrink to 20 minutes. A cancelled habit becomes zero.

Key takeaways

• Weekly trio contact beats quarterly research projects.
• Write the kill criterion on Thursday. Decide on Friday with numbers.
• A week with no production feature can still save a quarter of build.

Further learning

Teresa Torres at Lean Product Meetup on why weekly cadence beats project-based research.

Back to the map in continuous discovery.