3 ms·
I think you're describing biases in your sampling methodology. When folks do surveys do political polls, for example, the worst & lowest quality polls are basi
by flaviut 4y ago
I think you're describing biases in your sampling methodology.
When folks do surveys do political polls, for example, the worst & lowest quality polls are basically facebook polls because individuals select themselves by clicking an ad.
Higher quality polls work by having a representative sample of the population, and getting that is very difficult & time consuming.
In software, I think you have two options here:
- it is practical to capture analytics from about 100% of the population, so if it's possible to analyze this data to determine if your software works well, great!
- you can work hard to get feedback from a representative cohort. not sure how to do this... gift cards, swag, "you're part of the cool users club" communications? But be sure you're the one doing the sampling, and vet users' demographics (whatever that means for you) yourself.
- Closi 4y agoYes, both suffer from self-selection and also survivorship bias, however these steps further worsen that. It also depends what your goals as a project are - if you just want to build better software for your existing power-users, then great. If you want to become the next blender, you need to understand why people are using Maya/3DS Max instead (or in this example, why some people try to migrate to FreeCAD and fall back to using AutoCAD). It looks like the author was following this as much as possible though - i.e. not just focussing on some workarounds that a power-user had and focussing on how users should be able to do things easily without a lot of knowledge. The worlds most-used applications lower their barrier to entry as much as possible.