Most employed builders do the same expensive thing: they spend three months of evenings shipping a side project, then discover nobody wanted it. The code was fine. The screenshots looked sharp. The problem was never real enough for anyone to pay, switch, or even reply.
You do not need a survey panel or a Twitter poll to avoid that. You need five real conversations before you build. Not five “what do you think of my idea?” pitches. Five structured chats with people who already feel the pain—people who will tell you what they pay for today, what they hate about it, and whether they would try something else.
This is a nights-and-weekends playbook. Who to ask. A 20-minute script. A simple sheet for answers. Go / pivot / stop signals after n=5 (directional, not proof). A 14-day sprint that fits around a day job. No motivation speeches. No “just ship.” Just validation that does not require quitting anything.
What “five conversations” actually means
A conversation is a live talk—video, phone, or coffee—with one person who matches (or closely proxies) your intended customer. Twenty minutes. You ask about their world. You take notes. You do not demo your half-built app.
It is not:
A Google Form with twelve multiple-choice questions
A tweet asking “would you use this?”
Friends saying “sounds cool, I’d use that”
LinkedIn likes on your announcement post
You talking for fifteen minutes while they nod
Surveys and soft social signals are optional later. Before you write the first real feature, you want spoken specifics: the workaround they already use, how much it costs them (money or hours), and what would make them switch. That only shows up when you shut up and let them talk.
Five is enough to see a pattern and small enough to finish while you still have a job. It is directional, not statistical proof. Treat it that way.
Who to talk to (ICP proxies, not your group chat)
Ideal: people already paying for an ugly alternative—spreadsheets, agencies, bloated SaaS, manual ops, a VA. They have budget language. They have scars. They will not pretend a problem is interesting if it is not expensive.
Good proxies when you cannot reach the exact buyer yet:
Operators one step away — freelancers, agency PMs, community mods who serve your ICP daily
Recent switchers — people who just left Tool A for Tool B; they remember the decision
Cold-but-relevant — someone who posted about the pain in a Slack, Reddit, or niche forum in the last 90 days
Skip: your spouse (unless they are the buyer), your engineer friends who “would build that,” and anyone who only knows you as “the founder with the side project.” Nice people optimize for your feelings. Buyers optimize for their Tuesday.
How to get five names in one evening: list ten people from communities you already hang out in. Send a short ask: “I’m researching how [role] handles [job]—not selling anything. Got 20 minutes this week?” Expect ~30–50% yes if the ask is clear and non-pitchy. Book the first three before you overthink the list.
The 20-minute conversation script
Open with context, then get out of the way. Adapt the words; keep the order.
0:00–0:02 — Frame
“Thanks for the time. I’m trying to understand how people like you handle [job] today. I’m not pitching a product. Mind if I take notes?”
0:02–0:07 — Problem in their words
“Walk me through the last time [job] went wrong or took too long.” Follow with: “How often does that happen?” and “What does it cost you when it does?” Cost can be hours, missed revenue, stress, or actual dollars—capture whatever they name.
0:07–0:12 — Current workaround
“What do you use today?” Spreadsheet, tool, person, ignoring it. “What do you hate about that?” “What have you already tried that failed?” Listen for switching costs and “good enough” inertia.
0:12–0:17 — Willingness to pay / switch
Do not ask “Would you pay $29/mo for my idea?” Ask: “If something fixed [specific pain] without [annoying part of current workaround], what would you need to see before you’d try it?” Then: “Have you paid for anything in this area in the last year? Roughly what?” Past payment beats hypothetical enthusiasm.
0:17–0:20 — Closing signal
“Who else should I talk to who feels this weekly?” Referrals from people in pain are gold. If they offer to intro, write the name down before you hang up.
Optional last question if the chat is going well: “If I built a thin version that only did [one job], would you try it for two weeks and tell me what’s broken?” A yes with a calendar hold is stronger than a compliment.
How to log answers (one simple sheet)
Open a spreadsheet. One row per conversation. Columns that matter:
Who — role, company size if relevant, how you found them
Job / pain (their words) — quote fragments, not your summary
Frequency + cost — weekly? monthly? hours or $
Current workaround — tool + what they hate
Past spend — yes/no + rough amount
Switch trigger — what they’d need to see
Try-thin-version? — yes / soft / no
Referral offered? — yes/no
Your one-line takeaway — written within 10 minutes of hanging up
Do not wait until Sunday to “process” notes. Memory edits toward the story you wanted. Ten minutes after each call, fill the row while their phrasing is still sharp.
Go / pivot / stop after n=5 (with guardrails)
Five conversations will not prove product-market fit. They will tell you whether to spend the next month of evenings building—or not. Use blunt rules:
Go (build a thin slice) — At least 3/5 describe the same job in similar words, name a real workaround, and show past spend or clear willingness to try a thin version. At least 2 offer a referral or agree to test. Pain sounds frequent, not theoretical.
Pivot (change the job or ICP, not the color palette) — People care, but about a different job than you assumed. Or the pain is real for a narrower ICP. Rewrite the one-job sentence and run five more with the new target—do not “add features” to paper over a fuzzy job.
Stop (park the idea) — Compliments without workarounds. “Interesting” with no past spend and no try-commitment. Five different pains. Or everyone says the spreadsheet is annoying but “fine.” Annoying-but-fine does not fund nights and weekends.
Guardrails: if all five were friends-of-friends in your industry bubble, treat “go” as soft—get two more from colder channels. If someone says they’d pay but refuses a follow-up, count that as soft, not yes. And never let one enthusiastic call outweigh four shrugs.
When you later need honest product reactions (not idea validation), use the same bluntness you’d bring to brutal feedback from makers before a public launch—praise is cheap; specifics are not.
14-day nights-and-weekends sprint checklist
Goal: five conversations logged, a decision written down, and—if go—one thin experiment designed. Not a production app.
Day 1–2 (one evening) — Write one ICP sentence + one job sentence. List 15 outreach targets. Send 10 asks.
Day 3–4 — Book first three calls. Build the empty log sheet. Practice the script once out loud (yes, really).
Day 5–10 — Run conversations (lunch breaks, early mornings, one weekend block). Log each row the same day. Ask for referrals at the end of every call.
Day 11 — Hit five completed rows (or six if one was a misfit). Do not “wait for the perfect fifth.”
Day 12 — Score go / pivot / stop with the rules above. Write three sentences: decision, evidence, next thin step.
Day 13–14 — If go: define the thinnest test (landing + waitlist, concierge, or a manual “done-for-you” week)—not your dream architecture. If pivot: rewrite ICP/job and schedule the next five. If stop: archive the sheet and pick a different job. Do not “keep building in secret.”
Protect two evenings for talking, not coding. The whole point of this sprint is to earn the right to build later.
Common traps
Friends who are nice — They want you to feel supported. Buyers want their Tuesday fixed. Weight stranger/proxy calls higher.
Building while “researching” — Opening the repo mid-sprint is how you invent confirmation bias. Talk first; scaffold after the decision.
Treating LinkedIn likes as validation — Applause is not a workaround, a budget, or a try-commitment.
Pitching instead of listening — If you spent more than three minutes describing your idea, you ran a pitch, not a conversation.
n=1 hero story — One excited person is a case study candidate, not a market.
Survey theater — Fifty anonymous form responses with no follow-up path usually teach less than five live chats.
After go: talk to buyers, then earn the first dollar
A green light from five conversations means “build something thin and stay in contact with those people”—not “disappear for eight weekends.” Keep the same sheet. Add a column for “showed thin version / reaction.” When you are ready to find people who will actually pay while you keep the day job, use a concrete outreach loop like finding paying customers for a side project without quitting your job.
And when a tiny version works for a handful of people, resist the Product Hunt countdown clock. A quiet soft launch with real users beats a loud day with strangers—the same idea as soft-launching quietly before Product Hunt day.
Talk before you build
Side projects die from silence more often than from bad code. Five conversations will not guarantee success. They will stop you from spending a quarter of evenings on a problem nobody is trying to solve.
Write the job. Book the chats. Log the answers. Decide go, pivot, or stop—then act like you meant it.
When you are ready to put a thin version in front of other builders, browse what’s shipping on SideHunt—and show up with notes from real conversations, not just a polished landing page.