The engagement number that lies to you

A founder showed me their analytics last week. "Look at this," she said. "Average time on our pricing page is 4 minutes 12 seconds. People are really engaged."

I asked what the conversion rate was from that page. She paused. "Three percent."

Four minutes is a long time to stare at a pricing page before deciding not to buy. That's not engagement. That's confusion wearing engagement's mask.

We've trained ourselves to treat time-on-page and scroll depth as universally good signals. More time equals more interest. Deeper scroll equals higher intent. The logic feels obvious until you watch what actually happens.

The person who knows exactly what they want loads your pricing page, finds the plan that fits, and clicks in 45 seconds. The person who scrolls to the bottom three times, re-reads the feature comparison table, opens a second tab to check a competitor — that person is working hard because something isn't clear. They might convert anyway. But the effort itself is a cost you're imposing on them, and many won't pay it.

AI tools have made this worse, not better. They're remarkably good at generating engagement metrics that look healthy while the underlying behavior is stuck. Chatbots that keep users in conversation loops. Recommendation engines that surface more content without surfacing the right content. Personalization that increases clicks per session while decreasing satisfaction per click.

The problem is that engagement has become a proxy for value, and proxies drift. When a metric becomes a target, it stops being a good metric. We optimize for time spent, so we make things stickier. We optimize for scroll depth, so we bury the answer below the fold. The dashboard improves. The experience doesn't.

What would it look like to measure the opposite? Not how long someone stayed, but how quickly they got what they needed. Not how many pages they visited, but whether the first page was sufficient. Not engagement, but resolution.

This is harder to instrument. "Time to clarity" isn't a standard GA4 event. But you can approximate it. Track when someone stops scrolling — not because they left, but because they found it. Watch for the session where someone hits one page, stays 90 seconds, and converts. That's the ideal you're designing toward, not the 4-minute wandering session.

There's a counter-argument worth acknowledging: some products genuinely require exploration. A complex B2B tool, a considered purchase, a comparison between plans — these take time, and that time can be valuable. The difference is whether the time is spent learning or spent searching. Learning is additive. Searching is a tax.

The founder I mentioned ended up rewriting her pricing page. She cut the feature grid from 12 rows to 5. She moved the most-asked question to the top. Average time-on-page dropped to 1 minute 40 seconds. Conversion went to 6 percent.

Her engagement score got worse. Her business got better. Those two facts can coexist, and often do.