Blog posts with no conversions: keep them for SEO or prune them?

Our blog pulls steady informational traffic, but almost none of those visitors convert on the last click, and someone wants to cut anything that doesn't produce leads in GA4. I think that's the wrong test. A post answering "how much does X cost" often reaches people who are months from buying. They come back later through a branded search or the service page, so last click credits the branded visit and makes the blog look dead.

Before pruning, I'd look at which posts assist conversions, which still rank in the top 10 for money-adjacent queries, and which have lost clicks over the last 12 months. Thin posts, duplicates and anything competing with a service page are fair game to merge or remove. The rest I'd keep. How do you decide when a post has earned its place?

Agreed on the assist point. The detail I'd add is path length, not just whether a post shows up in assisted conversions. A post that sits two or three touches before a branded search is doing real work. A post that's the only touch before a conversion on a different day is more likely the reason someone found you at all. Both earn a place, but they justify different things: the first is a nurture page, the second is an acquisition page. In GA4 the Conversion paths report shows this if you filter to the conversion you care about. I'd sort posts into those two groups before deciding anything about the thin ones.

It comes down to how long your buying cycle runs compared to GA4's lookback window. If people usually take three months to buy and your window is 30 days, the two-touch posts will look like acquisition pages when they're really nurture, so the sorting goes wrong.

Agreed on the lookback window, and I'd say it comes down to conversion volume per post. With a thin funnel, a 30-day window can make a post look like pure acquisition when it's really nurture. But that only shows up if you have enough conversions for the paths report to mean anything.

Elena, "wait for enough data" can easily turn into never deciding, so here's a rule of thumb for the low-volume case. Below roughly 30 conversions per post in the window, one customer can flip the path length, so I wouldn't sort on paths at all. Sort on query intent instead. A "how much does X cost" post is nurture even with few visible assists, since that searcher is already in buying mode. A "what is X" post is acquisition, so judge it by rankings and clicks, not conversions. Revisit once the numbers are bigger.

Check the lookback setting first. GA4's attribution settings offer 30 or 90 days for conversion lookback, and if your cycle runs around 60 days, the 30-day view will misfile a lot of posts. Run the review at 90 days, then compare with 30 to see which posts move.

Falling back to query intent is a weak substitute. "How much does X cost" only tells you what someone typed, not where they are in the decision. Students, competitors and people checking prices for a friend all search that way, so sorting by intent means filing posts by assumption and calling it measurement.

With low volume, pool posts by topic cluster instead. Look at all the pricing posts together, or all the comparison posts. That often gets you past 30 conversions, and you can still check whether the individual posts agree with the group.

Pooling by cluster is a solid fix, but it works best if you pool the conversion side too. Pick one conversion event, like a form submit, and count only conversions where the first or last touch lands inside the cluster. Otherwise a pricing cluster ends up absorbing conversions that really came from a branded search.

Defining first touch by landing page has a gap. It only catches visitors whose session started on a cluster post. Someone who lands on the homepage, reads a pricing article two days later and then fills in the form never shows up.

A better approach is to build the cluster as a page group in GA4 and check whether any touchpoint in the path hits it. And when you add up across posts, dedupe the converters, otherwise someone who read three pricing articles gets counted three times.