GA4 engagement rate to judge blog posts: useful or just noise?

Has anyone made GA4 engagement rate the deciding metric for blog posts? An engaged session is one that lasts 10 seconds or more, has a conversion, or has 2+ page views, divided by all sessions. It sounds clean, but it swings a lot with page load speed, ad landing setups and how long people sit on a page before the timer even matters. Say a how-to guide sits at 35% and a comparison post at 70%. The how-to might be the one that ranks, earns links and sends people to the product page later. So I'd use it as a sanity check for broken pages, not as a signal for which content to invest in.

Mostly agree, but two details need fixing before anyone builds a dashboard on this. In GA4 the conversion condition is now called a key event, and the time condition is strictly longer than 10 seconds, not 10 or more. Small, but it matters if you're comparing against old exports. The more practical issue is that engagement time only counts while the tab is in focus, so a reader who opens a how-to, switches tabs for a minute, then comes back to finish will look like a bounce even though they read the whole thing.

Because of that, I wouldn't read the raw rate at all without segmenting it. Take a dental clinic blog: organic visitors reading a "how long do fillings last" guide and paid visitors landing from a retargeting ad will look nothing alike. Split the rate by landing page and by source first, then compare like with like. After that, check whether the low-engagement posts still produce key events a few pages later, which is closer to what you actually want to know.

Agreed on the focus problem. Short-answer posts get penalised too, though. Someone who lands on "how long do fillings last," gets the answer in six seconds and leaves counts as non-engaged, even though the page did exactly what it was supposed to. I'd check those pages against search impressions and downstream key events before cutting any of them.

Fair, and it gets worse when you report it to a client. A blended engagement rate makes the six-second answer page look like the weakest post on the site, and sooner or later someone will ask to cut it. I'd split posts into quick-answer and long-form buckets from the start and set a separate benchmark for each. That way the short pages get judged on what they're for: search impressions, clicks from the results page, and key events later in the funnel.