Where testers come from, how to recruit at the volume you need, and how to keep them opted in and active for the whole testing window.
Recruiting splits into filling a roster and keeping it filled. The first is a supply problem and gets all the attention. The second is where the campaigns are actually lost.
Most teams end up with a mix of these. Each behaves differently once the invitations are out, which is worth knowing before you depend on one of them.
Colleagues, clients and existing users. The highest opt-in rate and usually the most useful feedback, because they already know what the app is meant to do.
BehaviourFollow through reliably, but are rarely enough on their own, and the same few people get asked every release.
Places where developers trade opted-in installs. Fast to fill a roster from, and the fastest way to reach a number before a deadline.
BehaviourParticipation is transactional. Expect the minimum activity needed to count, and expect some to leave the track once the favour is returned.
Predictable numbers and quick to assemble, which makes them attractive when a window is already running late.
BehaviourActivity levels vary a lot between providers. Whoever supplies the testers, each one still has to opt in with a real account.
Newsletters, Discord servers, beta lists, social followings. Slower to convert than a pool and harder to predict.
BehaviourProduce testers with an actual reason to keep the app installed, so they survive the length of the window more often.
One roster, however the testers arrived
Every tester travels the same line. The question is how many are still on it when the window closes.
Before the invite
Start from the required tester count and add headroom. A roster sized exactly to the requirement has no tolerance for the opt-ins that will not happen.
Lost hereNo headroom. One tester dropping out puts the campaign below the threshold, and replacing them late risks interrupting the continuous period.
Stage 01
The address is on the roster and the invitation has gone out. This is the only stage where the number looks healthy and means very little.
Lost hereWrong or unused addresses. A typo, a work address the tester never checks, or an address already on the list under someone else's name.
Stage 02
The invited account opened the link and accepted. Confirmed against the roster rather than inferred from someone saying they had done it.
Lost hereSigned in with a different account. A willing tester on the wrong Google account fills nothing, and looks identical to success in a reply thread.
Stage 03
The tester has opened the build recently enough to count as participating. This is the number worth acting on during the window.
Lost hereQuiet dropout. Nobody announces it. The interruption is only visible if activity is checked through the window instead of at the end.
At close
The intersection of everyone you recruited and everyone who stayed. This is the figure the campaign record reports, and the one an application has to stand behind.
Lost hereA number you cannot defend, because the days it dipped were never recorded while it mattered.
Recruitment is the easy half. These are the habits that keep the number true.
Collapsing them is the most common way a campaign looks healthier than it is. Report the opt-in figure, act on the activity figure.
A tester who opted in and then stopped opening the app is the failure mode nobody announces. Checking weekly is enough to catch it in time.
Decide how many quiet days trigger a nudge and how many trigger a replacement, so the call is not being made under deadline pressure.
Anyone who stays opted in does not have to be recruited again, which is worth protecting even after a campaign closes.
Four messages cover a whole campaign. Anything more tends to be ignored.
All the sourcing in the world converges on one screen in your account. Everything before this point is supply; this is the list that decides your numbers.
Send us the app and the number of testers you need. We will build the roster, chase the opt-ins and tell you early if any slots are not being claimed.
