What Twitter's decline teaches founders about product metrics

Executive overview

Optimising for a single engagement metric — time spent, likes, video views — quietly destroys product quality. Twitter's shift to an algorithmic feed increased engagement while degrading user satisfaction and advertiser trust.

Every major social network follows the same arc: tight community, then growth, then algorithmic junk to prop up ad revenue. The only defence is a founder or product leader with the authority to define what the product is for and hold that line.

Engagement metrics measure activity, not value — and optimising for one number turns your product into an engagement farm.

The algorithmic feed trap

• Twitter's shift from a chronological feed to an algorithmic one mirrors TikTok's "show everything, let the algorithm decide" model.
• Users see more content, but content they didn't ask for: fistfight videos, political outrage, random junk.
• Engagement metrics go up; user satisfaction and long-term retention go down.
• Tools to fix the feed exist (mute, "not interested", curated lists) but are buried — when users need a Reddit thread to operate your product, that's a design failure.
• The business model matters: ad-funded products are structurally incentivised to maximise time-on-site, even at the cost of quality.

The social network decay pattern

• Facebook's newsfeed was the right move — it surfaced content users had chosen to follow.
• Over time, the network expanded beyond close ties; engagement-driven changes diluted the feed.
• Instagram repeated the cycle: early adopters, then mass market, then algorithmically injected content replacing friends' photos.
• TikTok skipped straight to the end state: pure algorithmic optimisation with no social graph constraint.
• Human social communities are bounded by Dunbar's number (~150 people); products that respect that constraint are structurally harder to monetise via ads.

Where X went wrong beyond the feed

• The blue checkmark conflated identity verification with paid subscription, destroying its credibility signal.
• Grok embedded inside Twitter is a product-scope mismatch — users don't understand why it's there.
• The rebrand to X discarded the verb ("to tweet") — one of the strongest brand assets a consumer product can earn.
• Renaming before the vision is clear does the opposite of what a name change should do.
• Taste doesn't transfer across domains: engineering excellence at Tesla and SpaceX doesn't translate to consumer social product decisions.

What Twitter still does well

• Breaking news remains uniquely valuable: first-person video surfaces faster than any media organisation.
• Community Notes has meaningfully reduced misinformation.
• Curated lists offer a high-quality, ad-free feed — but are nearly impossible to discover.

Advice for founders on product and metrics

• Define what your product is for and who it's for before hiring your first PM.
• Identify the "good states" and the "bad states" a user can end up in — engagement metrics won't distinguish them.
• A single optimisation metric will eventually consume everything else; it should be one input, not the goal.
• Founder CEOs have a structural advantage: the moral authority to hold a product vision against internal pressure.
• On naming: pick something easy to say, easy to spell, and ideally connected to the product — but meaning follows greatness, not the other way around.
• If you need to rename, do it early; the longer you wait, the more brand equity you forfeit.