Decide by value or a path to value, not by a traffic threshold. Keep any page that earns traffic, ranks for something, attracts backlinks, genuinely serves users, or supports a topic cluster as a hub or supporting piece. Prune or improve the pages that are thin, duplicate, off-topic, or generating no traffic and serving no purpose, with no realistic path to becoming any of those things. The single question that sorts every page is whether it has value now or a credible route to value soon.
This works better than a mechanical cut because traffic is only one of several ways a page can be worth keeping, and using it alone discards pages that matter. A page can get almost no visits while still ranking for a niche term, holding a backlink that strengthens the domain, or filling a needed slot in a content cluster so the hub above it makes sense. Cut it on traffic alone and you weaken things that were quietly working. By the same token, a page can have respectable traffic and still be a duplicate or off-topic mistake worth removing or merging. Traffic is a signal, not the verdict.
So the deciding criteria are concrete. On the keep side: does it bring traffic, does it rank, does it earn links, does it help real users, does it support the cluster. On the prune-or-improve side: is it thin, is it duplicate, is it off-topic, does it have no traffic and no purpose, and crucially, is there a path to fixing that. A thin page on a valuable topic is often an improve, not a delete, because it has a path to value. A thin page on a topic you do not cover and never will is a delete. The pivot is has-value-or-a-path-to-it.
To run this, build a list of your pages and score each against those criteria rather than against a visit count. Mark the ones with current value to keep, the ones with a path to value to improve, and the ones with neither to prune. Let value and purpose drive the cut, and let traffic inform it rather than decide it.