How to run a beta program for a new SaaS, step by step
To run a beta program for a new SaaS, invite a small group of people who have the problem today, give them access behind a feature flag or an allow-list, ask the same few questions every week, and set an end condition before you start. Ten to thirty testers is plenty to find the biggest problems. The beta comes before the public launch and has its own end date.
This guide covers seven steps, from picking a goal to closing the beta. The facts about tools come from PostHog's and Apple's documentation and Paul Graham's essay on early users, all read on 14 November 2026. The numbers and rules of thumb are ours and are marked as such.
What a beta is for
A beta answers one question you can state in a sentence. Examples: "Can a new user connect Stripe and see their first report without help?" or "Do people come back in the second week?" If you cannot write the question down, you will collect opinions instead of evidence.
Paul Graham's essay Do Things that Don't Scale makes the case for recruiting early users by hand. He writes that you cannot wait for users to come to you, and he describes the "Collison installation": when someone agreed to try Stripe, the founders asked for their laptop and set it up on the spot. Graham contrasts that with founders who ask "Will you try our beta?" and then say "we'll send you a link." Only the first approach guarantees that you watch someone use the product.
If you have not found those first people yet, how to get your first 100 users covers where to look.
The seven steps
- Write the beta question and an end condition. Our suggestion: pick a date or a number, such as "four weeks, or when 15 testers have finished onboarding, whichever comes first."
- Choose who to invite. Take people who already work around the problem. Graham notes that for B2B products the best early adopters are usually other startups, because they have not made all their tool choices yet.
- Gate access. Use a feature flag, an allow-list of emails, or an invite code. You want to be able to turn one person on or off without a deploy.
- Onboard each tester personally. Book 20 minutes, watch them sign up, and say nothing unless they are stuck for a minute or more. Write down every place they hesitate.
- Ask the same three questions weekly. Our set: What did you try to do? What got in the way? What would you miss if it disappeared?
- Fix the top item between check-ins and tell the testers you did. Testers who see their feedback acted on tend to keep replying (our experience, not a measured figure).
- End on your condition. Decide who graduates to paid, what you cut, and what the launch message will say.
Gate access with a feature flag
For a web product, the simplest gate is a feature flag with an opt-in list. PostHog's early access feature management is one example. Its documentation says you can run beta programs and waitlists with it, and that each feature moves through five stages: Draft, Concept, Alpha, Beta and General availability. Users opted in at the Alpha, Beta and General availability stages have the linked flag enabled. At the Concept stage a user can register interest, but the flag stays off, which makes it a cheap way to measure demand before you build.
Two details from that page are worth knowing. Opting in or out is an overriding condition on the flag, so it beats any other release rule. And when you promote a feature to General availability, there is a checkbox to include users who previously opted out, so decide in advance whether you want that. The page also notes that early access management works with the JavaScript Web SDK and does not support Groups, which matters if your accounts are organizations rather than individual users.
You can do the same with a plain database column such as beta_access and a check in your auth middleware. The tool matters less than being able to switch access per person.
If your beta is a mobile app
Apple's TestFlight overview gives the limits. You can invite up to 10,000 external testers and up to 100 internal testers who have App Store Connect access. The first build you add to a group for external testing goes to App Review, and later builds may not need a full review. A build can be tested for up to 90 days, after which it becomes unavailable, so your beta end condition should fit inside that window or you need to upload a fresh build.
What to track
Keep a simple sheet with one row per tester. Our suggested columns: invite date, first session, reached the main outcome (yes or no), last session, feedback notes, willing to pay (yes, no, unsure). After two weeks, look at the yes and no counts for the main outcome. If fewer than half of your testers get there, fix onboarding before you invite anyone else. That threshold is a rule of thumb, not a standard.
Graham also advises that you take extraordinary measures to make early users happy, and says that the feedback from engaging directly with your earliest users will be the best you ever get.
Turn the beta into your first customers
A beta ends with a decision for each tester. Offer the people who reached the main outcome a paid plan, and tell them what the price will be before the beta closes. The first 10 paying customers guide shows how to take payment without building billing first. Testers who will not pay can still tell you why, which is useful data.
When you are ready to be found by people outside your own list, a public launch comes next. Show HN vs Product Hunt compares two common places. Dealpuffin is another route if your product uses Stripe, Paddle or Polar: listing is free and you choose which coupons to publish.
Frequently asked questions
How many testers do I need for a SaaS beta?
There is no fixed number. Ten to thirty hand-picked testers is a practical range for a one-person team, because you can onboard each of them personally. Our reasoning is that you can only learn from people you can watch or talk to.
Should a beta be free?
It can be, but say so and say when it ends. Some founders charge from day one to test willingness to pay. Either way, tell testers the future price before the beta closes.
Open beta or closed beta?
A closed beta suits a first beta, since you control who is in and can talk to everyone. An open beta, where anyone can opt in, fits later stages when you want volume. PostHog's early access feature lets users opt in themselves, which is the open version.
How long should a beta last?
Pick a length before you start. Four to eight weeks is our suggestion for a first beta, and on iOS a TestFlight build is available for at most 90 days.