Something websites (*cough* Tumblr *cough*) need to learn is that what retains an audience isn't an abundance of new bells and whistles to play with, it's a coherent experience overall.
When someone joins a website, you don't need to grab their attention and hold it. They're already testing the waters. They've agreed to sign up. You've won on that front, and they're there for something specific you already have that they're hoping works well.
What drives them away is frustration.
Frustration, frustration, frustration.
Learning curves are going to be a part of any new website experience; they're something the user comes to terms with, in their own time. But broken or bad features are going to make them jump sites.
On top of that, constantly adding new features makes them feel like all the hard work they've put in to learning what you have isn't worth it; your website looks unstable and your staff looks incompetent, because it gives the impression that you don't know what you're doing.
You are floundering. It makes your new users nervous. It makes your old users hesitate to bring anyone else on board. And why should they? Why should they put effort into it if you're going to throw that effort away next Tuesday? Why get used to a UI that you're not going to bother to keep? Why customize anything if you're going to whittle that customization away?
Between that and the broken, unattended features of this site—the tag organization failing, the inability to look up posts word-for-word, the video player either refusing to play or yanking you to the top of the dashboard, images taking forever to load, advertisements blaring at full volume when you scroll past, you have your problem.
You have the reason why your numbers are failing.
It's not that you're not interesting enough.
It's not that you're too difficult to understand.
It's that you aren't improving what you have, yet you keep adding more half-broken things and unwanted copycat features to the pile.
It's that you're losing your identity in pursuit of a hypothetical perfect customer.
It's that you are actively telling your user base that you prefer those hypothetical customers over them. And your user base, your real people who make you happen, are smart enough to know where your priorities lie.
The bulk of this post talks about Tumblr, but other sites have gone the same way. Twitter is dead and its corpse is decaying in the street. Reddit has sabotaged any trust its users had in its management. If you'd like a really old example—I used to use Fanfiction Net. It's not the most intuitive website in the world, but it was the first one I called home.
I used it to host my works. The adware now on it makes it a hassle to navigate. The bots make comment sections and private messages a dread rather than a joy. So I moved on.
I also used to use it to collaborate on stories with my now-roommate. The message limit was 300 a day. When you're writing dialogue between characters, that's nothing.
So I moved on. We started messaging on Facebook. It was better, it didn't have a limit. But then I learned Discord existed, and I could edit messages, make dedicated channels, etc. So I moved on from Facebook to Discord. And Discord had a steep learning curve, especially if you're trying to make your own server rather than contribute to one. But, most importantly, the payoff was worth it.
If Discord changed its layout every other month while I was learning it, and broke how its reactions worked, and kept shifting what it meant to create a channel? If it opted me into servers I didn't sign up for, in hopes of engagement? If its text never formatted correctly, or its search function only went back a day or two?
I would have gone right back to Facebook. Even if it's a more basic experience, basic is always preferable to unstable.
Figure out what you want, websites.
Slow growth, or a gamble?
You're paying for your magic slot machine in users.
