Public launch week feels like the finish line. Soft launch is where you find out whether anyone can actually get value without you sitting next to them.
A soft launch is not a quieter version of the same hype post. It is a short, controlled period where real users run real workflows, you watch where they stall, and you fix the product before strangers vote on a listing. If you are picking a Monday start on a weekly platform like IndieHunt, that prep is what separates a spike of curiosity from a week that creates customers.
This playbook is for solo founders and indie hackers shipping AI tools who want evidence before applause.
Soft launch vs waitlist vs public launch
These three get mixed up. Keep them separate:
Waitlist collects intent and contact. People have not used the product yet.
Soft launch puts a limited group into the product under real conditions. You learn activation, pricing reactions, and support load.
Public launch week amplifies what already works. It is a distribution spike, not a product discovery process.
If you skip soft launch, your public week becomes unpaid QA with an audience watching. Fix bugs in private. Announce in public.
Define what “soft launch ready” means in one sentence
Write this before you invite anyone:
In [N days], [role] should be able to [core job] without me, and [one measurable outcome] should happen for at least [X of Y] users.
Examples:
In 10 days, freelance researchers should generate a cited brief without a call, and at least 12 of 30 invited users should finish one brief.
In 14 days, solo SaaS founders should connect their Notion and publish a weekly update, and at least 8 of 25 should complete two publishes.
In 7 days, agency ops leads should import a client list and send one status email, and at least 10 of 20 should send it without Slacking you.
If you cannot fill that sentence, you are not soft-launching. You are still building.
Recruit 20–40 real users, not fans
Fans cheer. Users get stuck. You need stuck people.
Aim for 20–40 invites. Fewer than 15 and one noisy user skews everything. More than 50 and you drown in support before you can ship fixes.
Who to invite
People who already described the pain in a community thread or DM
Past waitlist members who replied to nurture emails (not silent signups)
Warm peers in the same role as your ICP
Two or three slightly wrong-fit users on purpose — they reveal confusing copy fast
Who to skip for now
Anyone who only wants “early access vibes”
Competitors fishing for screenshots
Friends who will be polite instead of honest
Invite message that works:
I’m opening a private soft launch for [ICP] for [N] days. Goal: you try [core job] and tell me where it breaks. No public posts required. If useful, I’ll set you up this week — interested?
Cap seats. Scarcity is fine when it is real.
Remove the product features that create fake success
Soft launch fails when the product makes busywork look like progress.
Strip the path to one primary job:
One signup path
One “first win” screen
One obvious next action after the win
Kill optional settings from the first-run flow
For AI products specifically:
Default the model and prompts. Do not make users configure temperature on day one.
Show an example input and a good output on the empty state.
Cap runaway costs (token limits, rate limits) so a curious tester cannot burn your bill — or theirs.
Log failures with a human-readable reason in the UI, not a silent spinner.
Your soft launch metric is not “they clicked around.” It is “they completed the job you promised.”
Instrument five numbers before day one
You do not need a full analytics suite. You need five counters you check daily:
Invited — seats sent
Activated — finished signup and reached the main workspace
First win — completed the core job once
Repeat win — completed it again within 7 days
Blocker replies — messages that name a concrete stuck point
Ignore vanity during soft launch: social shares, upvote fantasies, “this is cool” without a finished job.
A simple sheet is enough. If first-win rate is under ~40% of activated users by day 5, pause invites and fix the onboarding path.
The 10-day soft-launch cadence
Copy this rhythm. Shorten to 7 days only if the product is tiny and support load is light.
Day 0 — Invite + setup
Send invites in two batches (morning and next-day) so you are not slammed. Offer a 15-minute setup slot or a Loom. Write the first-win definition in the welcome note.
Day 1–2 — Watch silently, then nudge
Check who activated. Personal nudge to anyone who signed up but did not finish the core job:
Saw you got in — stuck on [likely step]? I can jump in for 10 minutes or send a 90-second Loom.
Day 3 — Fix the top two blockers
Ship the two changes that would have unblocked the most people. Announce briefly to the soft-launch group. Do not add three new features.
Day 5 — Midpoint review
Count first wins and repeat wins. Interview 3 users who won and 3 who stalled. Ask the same three questions:
What were you trying to finish?
Where did you almost quit?
What would make this worth paying for next month?
Day 7 — Pricing pulse (optional but useful)
Share a draft price and ask for a gut check, not a commitment:
If this stayed as-is, would $X/month feel fair, high, or confusing — and why?
You are collecting language for the public listing, not closing deals yet (unless someone insists).
Day 10 — Go / no-go for public week
Decide with rules, not vibes. See the checklist below. If you are a no-go, extend soft launch instead of locking a Monday slot out of calendar anxiety.
Support rules that keep you sane
Soft launch is intentionally unscalable. Guardrails still matter.
Office hours: one daily window (for example 30 minutes) instead of 24/7 Slack
Bug labels: blocker / annoying / nice-to-have — only blockers pause the clock
Public channel optional; private DMs for account-specific issues
Never argue with feedback in the moment — log it, thank them, ship or reject later
If one user needs a custom script to succeed, that is a product smell. Either productize the script or admit that ICP is wrong.
Go / no-go checklist before you book a public launch week
Say yes to a public week only if most of these are true:
At least 10 users completed the first win without a live call
At least 5 completed a repeat win in the same week
The top blocker from day 3 is fixed and re-tested
You can explain the product in one sentence a stranger understands
You know the first email you will send launch-week visitors
Support load is declining, not climbing, as new invites activate
You have 2–3 short quotes or screenshots you have permission to share
If three or more boxes are empty, keep soft launching. A weak public week teaches the wrong lesson: “launches don’t work,” when the real lesson was “the product was not ready to be alone with strangers.”
When the boxes are mostly checked, schedule the week on IndieHunt (or your chosen platform) with a listing that describes the job users already finished — not the roadmap you hope to ship during the spike. Sharing a build update on Indie Hackers can recruit a few more soft-launch testers if you ask for critique, not votes.
What to put on the public listing after a real soft launch
Your listing should steal language from soft-launch chats:
The pain words users typed
The outcome they finished
The objection you already killed in the product
Who it is not for (soft launch wrong-fit users give you this for free)
Do not invent social proof. Use the quiet proof you earned: “Used in private by N [role] who completed [job] in the last two weeks.”
Soft launch is the dress rehearsal; public week is opening night
Treat the days before your public launch week as a stress test, not a teaser campaign.
Recruit users who will get stuck. Measure first wins, not compliments. Ship the blockers. Decide with a checklist. Then go public when the product can stand without you hovering.
That is how an AI indie launch stops being a lottery and starts being a system.
