I was at StumbleUpon at shutdown and for a few years before. I'm creating a single-purpose account to avoid doxxing myself on my main account. Here's a non-exhaustive list of things that killed SU.
SU relied on individual human curators to discover worthy new web pages instead of trying to crawl the web in bulk. Our most passionate users, the people who would submit great new URLs, tag URLs, rate heavily, and curate good collections, were heavy users of our XUL based Firefox extension. We were never able to build an extension as featureful post-XUL and that hurt new content acquisition and categorization.
The linked content we served to users on the web had to be iframeable. That described most of the web in the early years of SU. By the time of the shutdown, most new sites and even many established sites couldn't be displayed in iframes any more. It would have been technically possible to replicate most of the iframe experience without iframes by serving up our own crawled copies of pages, but then we would have been infringing copyright. Iframes allowed SU to show a third party page without actually copying it.
Over time, more sites disallowed crawling via robots.txt or by thwarting them without prior declaration by rejecting "crawler-looking" requests that didn't come from Google.
Our mobile apps allowed us to serve content in a webview even if it wasn't iframeable, but mobile platforms inherently excluded a lot of the "weird and wonderful" stuff that people loved SU for. For example, our online games category contained mostly Flash games. Even most of the non-Flash games assumed either keyboard or mouse access; few of them worked well on a phone's touch screen. Similar problems applied to other interactive content (simulations, interactive visualizations). Take away the interactive content and SU started to look more like just another app for passive scrolling.
Web pages became increasingly encrusted with ads, nags, and tracking scripts. I was the final developer responsible for maintenance on our Android app. 95% of the performance complaints we received about the app were, in truth, complaints about the analytics and ads an individual recommended web page was loading. This was also the source of 95% of the complaints about ads that we received. SU monetized itself with interstitial ads between recommended web pages. Every ad running on a recommended page was put there by the site owner, not us. But negative reviews and customer emails blamed us for the battery-killing, device-heating, content-obscuring ads running on-page. The ad problem was much less severe for people who used the web site instead of mobile apps, since they were largely power users who had ad blockers installed, but see previous problems about iframes and Firefox plugins.
Consolidation toward siloed content platforms certainly didn't help either. But I think that SU would have been able to keep going until the present day if not for the problems mentioned above. Some of them seem like mere happenstances of history; I could imagine Flash living a lot longer if it hadn't been a notorious source of exploitable security holes. Other things, like long-form articles getting increasingly barnacled with ads, nags, analytics, and paywalls, seem like a more inevitable outcome of the collapse of old-media print revenue.
One thing I've idly pondered since SU shut down is just recommending weird-old-web content that's findable in the Wayback Machine yet missing from the modern internet. The beauty would be two-fold: it's likely to be cleaner content in the first place, and you can transform it (e.g. running Reader Mode over it server-side) without worrying about the site owners coming after you for copyright infringement. Or at least not worrying as much. It seems like marginalia_nu has managed to find a goodly amount of content that still lives so maybe resorting to the Wayback Machine isn't necessary at all.