For a bit of computer history, while it may sound quite far fetched nowadays, there were actually times when relaunches were generally welcomed and user numbers were expected to double rather than dropping. (Well, this was before UX.)
Edit: As an example, I suppose none of the much beloved and personality defining original Macintosh icons by Susan Kare could have come out of a modern UX work flow, nor the use of the Norwegian landmark sign as the marker of the command key. Or, where there any users that rejected the OS X Tiger and Snow Leopard UI, which applied quite radical changes to the previous Aqua UI?
Windows 95 gave you all kinds of stuff Windows 3.1 didn't have, Nintendo 64 vs SNES is another example.
Users want upgrades, not redesigns. If you give them a upgrade, no one cares about whitespace or rounded corners, that's why those things used to be an afterthought.
I know a few people who left Mac due to this change.
Win3.11 to Win95 springs to mind here.
In contrast, going from "good UX" to "better UX" is a whole different endeavor since you'll have to fight against the friction of people having to give up routines and flows that they have been sticking to for a long time. A thing which no amount of thoughtfulness and consideration of a redesign will be able to make up for – that is, initially.
For me this was the starting point.
After or around the same time someone from Gnome (or Red Hat?) realized (at least this is how I remember their marketing) humans think spatially - and used that observation to decide that everytime I click on a folder in Nautilus it should open in a new window. (Yes, it becomes a mess really quick.)
Also the insistence on removing every configuration option.
Same goes for Unity: my best explanation is someone heard Mac OS had great UX and decided to copy it wholesale: Move controls to the left. Change Alt-tab. Introduce dock.
Common for all these "improvements" is that nobody asked us users, and nobody cared to make it easy to go back.
I now have an acquired distrust of ux designers (and I am really happy whenever I meet good ones which thankfully happens often where I work.)
- Ribbon has mostly been fixed now, and is mostly pleasant to use even if the File menu is still jarring.
- Nautilus was fixed in the next major version.
- Unity was never fixed. I hear some people like it but I guess we are a good bunch of old Ubuntu users who cannot stand it. More importantly it is hard for me to see any reason besides arrogance for why there couldn't an option somewhere to get back to normal alt-tab.
The way I see it, arranging things in space works well when you have enough space to fit these things in; once you have to scroll or zoom, the utility rapidly degrades with the number of excess stuff. On top of that, remembering where things are works when there are couple things. Not when there might be couple thousand.
For comparison, it's worth thinking about two other things human beings seem to understand and employ rather well:
1) Lists. Note that while you may put different lists in different places, you generally don't generally put each item on a separate post-it placed in a different location. List items belong together.
2) Hierarchies. That one is probably biological, and almost as fundamental as understanding 3D space. We inherently grok hierarchies - both abstract and social. We navigate them every day.
Unlike spatial arrangement, hierarchies and lists scale well. And so that's why a filesystem is a hierarchy, and it's usually rendered as either list, or - if you need to see multiple hierarchy levels at the same time - as a tree.
I at least felt Gnome 3 was about the same only it was possible to configure it.
My plan was to use the Pop OS edition since they actually had put some thought into shortcuts.
Another DE I actually enjoyed despite being initially skeptic was Elementary OS.
After KDE 5 became stable though I've stuck with thatm
The designer and execs only look at new screenshots of the design, and ignore usability. Maybe a few minutes of thought and review are given to how a user interacts with the site, but only to talk through a narrow set of interactions that the designer took the time to bother with.
Site keep getting less functional (reddit, linkedin, yelp, twitter) and slower (everything google). All these companies encourage short-term thinking, and incentivize whatever looks good on a powerpoint or checks a box on their year-end goals.
Interesting observation. I've had this very situation with mobile apps, where users base their decision on whether to download an app largely on the basis of having a quick look at the screenshots. Adding the proper functionality and improving discoverability inevitably leads to a design which simply isn't as slick as the minimalist version. The only solution I can think of is to have an "advanced" version of the UI which users can enable.
This. (I have high hopes for Apple, now that J. Ive is gone.) However, mockup screenshots had always been the medium of communicating designs to the client. But the more prominent role of screenshots may further the concentration on just a few paths of usage, as we may observe it in the common reduction of interface options – which may be fine for the few selected usage paths and may provide the base for a "cleaner" design (which should actually be a design job, not a content issue), but necessarily fails a number of use cases.
Another point may be made about an overly application of metrics. The issue here is that all metrics that may be actually useful are in the qualitative domain, which is rather labor intensive, complex and expensive. (Especially the process of transforming qualitative data into quantitative metrics.) So everyone is going for quantitive data, which is cheap and supposedly easy. That only few of those involved have a background in social research and would actually know what they are doing, doesn't especially help. So we may add data driven design to the list of possible issues.
(Much of this may be understood in the old dichotomy of "art or engineering". While the engineering approach is fine for managing exchangeable labor, it's failing in many other aspects. Meaning, what you get is an exchangeable product and chances are that, if you actually have a unique product, this approach will fail your needs – or, at least, the needs of your customers.)
Then I tried Gnome and it was superlative in comparison. Honestly, maybe it was because we had nearly nothing. Everything was new and magical.
Now we start with a high baseline. Everything is already pretty good. A redesign rarely gives me more.
For a striking example, have a look at the hitherto excellent Norwegian weather site's short term forecast here https://www.yr.no/place/Portugal/Lisboa/Lisbon/hour_by_hour.... then click through to what they're humbly bragging about changing it to https://www.yr.no/en/forecast/graph/2-2267057/Portugal/Lisbo...
I can only assume neither the management nor the "design" consultant actually use the result. Oh, and consultant reminds me. Accountability is probably a major factor here. In my perception, the people causing this sort of destruction don't suffer any negative consequences of them. Maybe they even get promoted or get a prestigious client to list on their CV.
Looking at your weather site example, this is precisely the kind of stupid design decisions I so hate about modern software. On the old site, all the data is in one place, and can be read without any interactions. You can just glance at it and learn what you need (which also means you can do that while holding a toddler in one hand, and eating your sandwich with another). Who on Earth thought that dumbing down the table and forcing you to manually switch between it and the graph (and no option to show both!) was a good idea? What's up with information moved to graph tooltips? And where has the link to download the PDF version went?
Your point about accountability is probably spot-on. These changes aren't making the website completely unusable, just strictly worse and more annoying than before. That's nowhere near enough to cross the threshold after which anyone on the vendor side would care. That, multiplied by the number of users, it amounts to a lot of man-units of annoyance, doesn't even register.