How to get feedback on a SaaS before launch
I did not add comment tags to Nukkadly for decoration. Launch boards fill up with "congrats" and the maker still does not know what to ship on Monday. The same sort works before you are public. Five people who already do the job, one task, four labels. Then change the product. Then show it.
Updated September 24, 2026
Why a public week is a bad first interview
A launch week is strangers with a few minutes. They will vote. Some will comment. Most of those comments, on every board I have used, are social. They are not a test of the first screen. If that screen already fails, you just paid for a week of people bouncing, and you will tell yourself the board was quiet.
So I made comments on Nukkadly pick a kind: useful, confusing, pricing, missing, plus a plain question if they actually have one. Empty congratulations without substance do not help anyone. That is a product decision, not a slogan. You can run the same four labels in a notebook with no account. You should, before you submit.
Five people, one task
- Write the job in one sentence a customer would say.
- Find five people who did that job in the last month. Not four founders and your cousin.
- Give each person that job. Stay quiet while they try it.
- Ask what they did last time, where they stopped, and what they expected.
- Label every note. Fix the repeating confusing item. Repeat with two new people.
Maker friends will review your type and your positioning. They already know how these products are supposed to work, so they will not fumble signup the way a buyer does. Keep one of them if you want a sharp pass on the copy. Do not let them be the whole sample.
Find the rest where the complaining already happens. A Slack about the job, the replies under a competitor changelog, a subreddit that is not about startups. Ask for twenty minutes to try a task. People say yes to that. They say no to "can I pick your brain?"
Watch the pause. Then ask about last Tuesday.
"Send last week's failed invoices to the finance lead" is a task. "Take a look around" is a tour you will narrate. If you rescue them at the first stuck click, you tested your coaching.
After they finish or give up, ask about the past. Rob Fitzpatrick wrote this down as The Mom Test. Talk about their life, not your idea. "Would you use this?" is a polite yes. "What did you do the last time invoices failed on a Friday?" is harder to fake. Follow it with where they stopped today.
The four labels we actually store
A doc full of quotes feels like progress. It is a pile. These are the same kinds a visitor picks on a Nukkadly product. Use them in the calls so the public week, if you do one later, is not a second language.
Confusing
They could not tell what to do, or they did the wrong thing with confidence. Fix the first screen. Do not add a feature to cover a sentence people cannot read.
Missing
They named a job the product does not do, and they were blocked. Wait for a second person to hit the same wall. One wish list is how you grow a settings page nobody opens.
Pricing
They asked what it costs, or they named a bill they already pay. Write down that tool and that number. Your pricing page should answer the comparison, not a round figure you guessed on a Sunday.
Useful
They described a job they would repeat, and they said when. "This would have saved me the Friday export" counts. "Love this" does not. Leave it out of the doc.
Would try is cheaper than would pay
On the live board, visitors can mark fits my use case, would try, and would pay as separate signals. In evergreen rank, would pay is weighted 1.2, would try 0.7, fits 0.35. I put the numbers on how ranking works because I got tired of boards that treat a curious click like a purchase. Someone who would try is being polite with their calendar. Someone who would pay is stacking you against a bill they already have. Do not set a price from the first group.
Those marks still are not the five calls. They are a check, later, on whether strangers who do not owe you a kind word lean the same way. How a launch week works is the loop after the product can be opened.
When to stop
Fix the confusing item that showed up more than once. Bring in two people who were not in the first round. If they get past that screen without you talking, and the notes rhyme, you are done with this round. More calls after that mostly repeat the early quotes with softer adjectives.
Silence counts. If they agreed to try it and never opened it, do not send a survey. The job is not urgent. A launch week will not make it urgent.
When a stranger can finish the task, pick where to show it. For a developer tool, start with the install path. That choice is where to launch a developer tool.
Common questions
How many people should I talk to before launching a SaaS?
Five people who did the job in the last month is enough to start. Stop when two new people get through the task without hitting the bug you just fixed, and the new notes repeat the old ones.
Are surveys a good way to get pre-launch feedback?
Later, to people who already used the product, when you are checking a price or a label. A survey will not show you where the first screen fails. You have to watch that.
What should I ask instead of do you like it?
What they did the last time the problem showed up, where they stopped in your product, and what they expected next. A future yes is cheap. Last Tuesday is not.
Can a Nukkadly launch week replace those conversations?
No. Tagged comments and would-pay marks only help after a stranger can finish a task. Use the five calls first. Use the board when you want people who do not already know you.
After the five calls
Submit when a stranger can finish the task. Tagged comments only help if people can use what you listed.