[latexr]

The post only mentions a few things:

> we’ll be able to gather only a small set of general events and interactions within our apps. Things like when you unlock the app, when you create a new item (but not its contents!), or when you use autofill (but not what sites you use it on!).

[ehPReth]

call me stupid; but I'm not sure how those numbers are helpful for them?

[selykg]

How are people creating new items? App, or extension?

How are people accessing items? App, quick access menu, extension, browser bookmark?

How are people changing passwords? In the app, using the password generator or not, in the extension with the password generator, in the browser using the injected UI?

The thing about 1Password is that it seems like it's simple, but under it all there's usually multiple ways to do the same thing. Using some telemetry they could easily see that only 2% of users are using this one particular feature, and cut it if it's not getting used. Or this fantastically useful feature is only getting 20% of users using it, maybe they need to introduce it to users in a better way. Etc.

At the end of the day, having this kind of data can make for better decisions. I'm not a fan of telemetry though. I'm honestly surprised the security team at 1Password agreed to this one as well.

[latexr]

> having this kind of data can make for better decisions.

It can also make for worse decisions.

2% of millions of users is still tens of thousands of people. Maybe that feature is terribly useful but only a handful know about it. Cutting it would be a mistake; it should be made more prominent.

Maybe the 20% feature is annoying and that’s why 80% of people actively avoid it. Giving it more prominence would be a mistake; it should be cut.

No amount of telemetry will tell you users are deeply unhappy with the move to an Electron app and the removal of local vaults. You only know that from direct feedback and speaking to them.

[filchermcurr]

That's the trap Firefox fell into, I think. Relying so much on telemetry that only tells a very small part of the story. So they cut features that are useful, but people didn't even know they existed. It seems like this data-driven approach should go: Collect data, form conclusion, try to increase awareness of feature, collect more data, form new conclusion, cut if awareness didn't pan out.