Prune a low-performing page only when it is dead weight with no path to value, and leave or fix it when it still serves someone. The pivot is exactly that: dead-weight-with-no-path versus salvageable. Pruning helps when a page is thin, duplicative, or off-topic, drags on the site’s overall quality signal, and you have confirmed it cannot realistically be improved into something useful. Removing that kind of page can lift the site by shrinking the share of low-quality content judged against you, which is the whole mechanism, not deletion for its own sake.
The confirmation step is what separates a smart prune from a reckless one. Before cutting, check whether the page actually has no path: does it get any traffic, does it rank for anything, does it answer a real query even poorly, and could a rewrite turn it into something worth keeping? A page that is merely underperforming today is not the same as a page that is structurally hopeless. The first is salvageable and the second is dead weight, and only the second belongs in the prune pile.
Plenty of low performers should stay. A page can earn few visits yet still serve users who land on it, hold valuable backlinks you would lose by deleting the URL, support internal structure, or sit one good rewrite away from working. Cutting those is self-harm dressed up as cleanup. The two failure modes sit on opposite poles: pruning aggressively in the belief that deleting pages mechanically boosts the site, and refusing to delete anything out of attachment. Both skip the only question that matters, which is whether the page has a path to value.
So audit each low performer against that single test before deciding. Confirm the genuinely dead pages, thin, duplicate, or off-topic with nothing to redeem them, and prune those. For everything that serves users, holds links, or could be fixed, keep it and improve it instead of cutting it.