I often wonder these days about the value of telemetry and A/B testing. I think many heavily used pieces of software can benefit from a BDFL-like decision maker instead, especially if the alternative is software specifically designed to use me as a guinea pig surreptitiously. I feel like, in the race for adoption, we suffer a form of the tyranny of the majority with software. Decision making based on principles instead of popularity is the only way to prevent harming some of us whose software preferences are in the minority.
It seems A/B testing provides clear numbers on what's most popular, but you can't quantify what's right.
Indeed. When A is shit and B is shit then the result of your A/B testing is going to be the least shit of the two. At the end of the day it's going to be shit either way.
I'm a firm believer that you should never ask your user to make a decision or look over their shoulder. Not once ever. You should listen to their complaints and ideas when they come to you, then build your strategy on that. Be reactive, not prescriptive. That empowers the user, shows respect and results in a satisfactory product that benefits the user which after all is the end game.
Telemetry invades the user's privacy. Feedback does not empower the user because the user expects a reaction from it which is unlikely. A/B testing results in churn for the user which does not show respect, merely that they are a test subject.
Microsoft as a fine example could learn a lot from listening to their users rather than steamroll ahead based on collected telemetry and feedback data.
A fine example: People didn't want UWP/metro and still don't today. I have yet to meet one person who uses that side of windows 10. They wanted shit that worked, was faster and kept out of their way and didn't wreck the workflow that they had invested years in learning or had someone experienced close at hand to help them with it. 90% of the userbase just installs chrome and does everything in there as well so that stuff just gets in the way.
> I have yet to meet one person who uses that side of windows 10.
I was going to say, "I do!", but then I realized that no, actually, I don't. I mean, I have a Win10 tablet PC which I often use in tablet mode, and I appreciate the Windows 10 UX. But what I most appreciate about that UX is... how well it works with regular Windows applications!
Because it turns out UWP/metro is just too dumb an interface. It's Android/iOS-level dumb (just with less apps). I'm split about Microsoft right now. On the one hand, I just don't understand why they're on the "dumb down everything" bandwagon. On the other hand, as long as they still support normal Windows interface and applications, I want to support them and wish them best, because Windows 10 is literally, honest-to-God, the best system for tablets that currently exists. Period. I don't want to have to move back to Android.
I find Windows trend to be quite odd, too. And that's really telling of an issue, because I'm still using my Windows Phone and desperately want more quality apps. However, when I move back to my desktop, I rarely use UWP apps--with the exception of a handful that sync with my phone, like podcasts or my budgeting app.
Even when I use my Surface, I rarely end up using the UWP apps because, as said, they're dumb. Even the Microsoft ones feel half-baked and lacking. I think Microsoft should have dumped more money into Research and tried to find a novel method for automatically handling different UI sizes without dumbing down the UI.
I like the idea of the Store, too, but I think instead of forcing all apps to this new interface, they could have created a new package format, like Mac's _.app_ directories, to allow one-click install distributions of classic Windows applications.
Strongly agree. I said that before back during Homebrew's telemetry fiasco - no, you don't need analytics in your desktop software. You especially don't need opt-out, active-by-default analytics. People made software totally fine without spying on everyone back in the day, and I'd hazard to say that it was often better software.
> I think many heavily used pieces of software can benefit from a BDFL-like decision maker instead, especially if the alternative is software specifically designed to use me as a guinea pig surreptitiously. I feel like, in the race for adoption, we suffer a form of the tyranny of the majority with software.
I’ve recently been feeling the same but couldn’t articulate it as well as you have.
This is one thing that the biggest open source projects seem to get wrong more often than right (trying to be all things to all people, all of the time). Management by consensus usually yields mediocre results in the commercial space, too. The BDFL model is a great compromise. Everyone gets a say, but not the final say, and there’s a consistent vision driving the project.
When projects get very popular, it’s hard to say “no” and keep that vision focussed. This seems to come more naturally to proprietary projects, at times resulting in a better overall experience for the narrower subset of users served.
There are advantages and disadvantages to community governance. It is easier for people to feel ownership and a governance stake. Projects tend to value backwards compat more consistently, rather than following the preference of an individual BDFL, since stakeholders who are injured by a compat break can have more say. This stability over time makes them safe to invest in.
Plugin based architectures tend to work well under such governance, as then the contributors are more insulated from each other and interfere less frequently.
No single governance mechanism is so much better that the others need to go away. Not every project should have community governance, not every project should have a BDFL. Not every project should even take outside contributions -- personal projects (especially licensed under CC0) can offer great value to the wider world without having to placate opinionated contributors.
It seems A/B testing provides clear numbers on what's most popular, but you can't quantify what's right.