3 ms·
If I can get a clarification, how well does the law of large numbers apply to a non-random sample? In other words, 10k paying users are self-selected into the
by Shamiq 18y ago
If I can get a clarification, how well does the law of large numbers apply to a non-random sample?
In other words, 10k paying users are self-selected into the service, which would include biases in measurements. Thus, depending on the variation of the user population, it could theretically be possile for small stimuli having large effect to a user base. If the users are similar, but are part of a niche, if that niche gets hit, the business could be wiped out.
- nostrademons 18y agoIt doesn't. If your sample is non-random, it's quite possible for you to lose your 17,333 users in short order, if some external force acts upon them all. That's basically what happened with the subprime crisis - a lot of supposedly "independent" homeowners all defaulted at once, which played havoc on the computer models for valuing CDOs. They all had the unknown common factor that the loans were based on permanent home appreciation, and so couldn't be sustained on income alone. So when house prices started declining, bye-bye subprime loans. How secure this makes your customers depends on why they're choosing your service. It's very unlikely that 37signals or FogBugz would lose all their customers at once (unless a great new competitor sprung up), because they're all separate, independent businesses with their own reasons for purchasing the product. It's fairly likely that Twitter or FaceBook could lose all their customers, if there was some reason for an exodus to start, because most of their users are only on the service because their friends are.