A smaller, sharper launch
A beta gives a limited group access to an unfinished product so you can learn from actual use. Pick one audience and one task. A busy invitation list is less useful than five people who repeatedly encounter the problem you want to solve.
What to observe
Watch a new user get from signup to the first useful result. Record where they stop, what they expected, and whether they return without prompting. Ask about their previous workaround before asking for feature ideas. Personal onboarding is useful early: Paul Graham’s essay makes the case for recruiting and helping initial users manually.
Set an exit condition
Choose a concrete test before inviting the cohort: for example, whether participants can complete the core task without your help. This is an illustrative test, not an industry benchmark. Fix the common blocker, repeat with new users, and widen access when the result is repeatable. Ask permission before publishing customer feedback or screenshots.