[ajdude]

> A few weeks ago, we conducted hour-long conversations with 10 of our users to dig deep into how you manage your preferences and configurations in Thunderbird desktop

Is only 10 people representative of the population of thunderbird users?

[Aachen]

Fwiw, even just going through your software with one user can give quite a few insights about what's not obvious about it. That's not at all to say you never need more, but very few open source projects do user research in the first place, being passion projects that just scratch the developer's/s' itch. More samples is always better, definitely at n=10, but I'd also not dismiss the results and benefits of doing it!

[InsideOutSanta]

Also, how many people you need depends heavily on whether the thing you're researching affects a lot of people. If settings has a problem that affects 40% of people, then a sample of 10 people would yield results representative of the whole population.

Sample size is a very weird, often kinda counter-intuitive topic.

[daneel_w]

If the crowd is diverse enough. Was it? The article doesn't reveal.

[angiolillo]

The video goes into slightly more depth, and at about 1:30 into the video they acknowledge that the participants were not representative and that they would like to conduct further research.

[angiolillo]

> Is only 10 people representative of the population of thunderbird users?

For very narrow studies it is possible to get representative data with fewer than a dozen interviews, but in this case it is explicitly not representative. In the video they mention that most of the participants have used Thunderbird for over a decade and follow release notes, development, and various forums closely, which to me suggests that they were recruited opportunistically rather than a random statistical sampling.

They do mention that they have plans to engage a larger audience in the future but that can be incredibly expensive. Even large organizations typically have to augment a small number of representative interviews with a large number of surveys and a very large set of user telemetry to properly weight interview feedback.

[mmooss]

It's a standard research technique. You can have 2,000 people answer an automated survey but you can't have hour-long conversations with them. Researchers in many fields would like a better solution for in-depth interviews.

[BeetleB]

Thunderbird has as many as 10 users?!

(I jest!)