Loop Navigation: The Frustration Signal You're Not Watching
Users bouncing between two pages 4+ times in 90 seconds aren't browsing. They're lost. Here's what loop navigation detects and why it matters.
Our /pricing page had a loop navigation score of 34 before we understood what that meant. Users were hitting pricing, going back to the homepage, returning to pricing, back to homepage again. Same 4 users. Same 90-second window. The pattern was so consistent it looked like a bot test.
It wasn't a bot. It was real people who couldn't figure out what they were buying.
What loop navigation actually is
The signal fires when a single user visits the same two pages 4 or more times within a 90-second window. Not browsing. Not comparing. A navigation pattern that means something on one of those pages isn't answering the question they came with.
The signal weight is 20 in the confusion score. Second highest after rage clicks (weight 25). It's weighted that heavily because it's almost never accidental. A single rage click could be a misfire. Four round trips between two pages in 90 seconds is a person who's frustrated and looking for something that isn't there.
Why it's worse than bounce
Bounce rate treats "left" as the outcome. Loop navigation treats "kept trying" as the signal.
A user who loops is more motivated than a user who bounced. They want something. They're willing to look for it. The fact that they loop instead of convert means the answer is somewhere on your site, they believe it is, and they can't find it.
I'd take that user over a bounced one any day. The looper is telling you something specific. The bouncer might have just been the wrong audience.
For most products, loop navigation concentrates on two page pairs: pricing-to-homepage and feature-page-to-pricing. Both patterns have the same cause. The user has a question the first page raised but didn't answer, and they think the second page might have the answer. It usually doesn't either, which is why they keep going back.
What triggers it
The loop detection watches for the pattern within a session, not across sessions. So it's four round trips from the same session in under 90 seconds.
The 90-second threshold is intentional. Longer gaps would capture casual browsing. Under 90 seconds means the user is actively navigating, not taking breaks. They're in a confused state right now, not slowly reading the site over 20 minutes.
The four-trip minimum filters out the user who accidentally hit back and re-navigated. Two trips: plausible accident. Four trips: something is wrong with the pages they're bouncing between.
What to do when you see it
The page pair matters. If it's pricing-to-homepage, users have a question your pricing page didn't answer. Common ones: what does this actually do, is there a free trial, what counts as a "session."
If it's feature-page-to-pricing, they're comparing what they get against what they pay. The feature list and the pricing table don't match up clearly enough for them to make the decision.
The fix is almost always information, not design. Add the thing they're looking for. A one-line answer to "what counts as a session" on the pricing page eliminated loop navigation on that pair for us completely.
What it's not usually is a design problem. Don't redesign the page because users are looping. Read what the loop is telling you: they believe the answer is on your site, they can't find it, and they're checking twice to make sure. Find the missing piece, add it, and the loops stop.
The co-occurrence signal
Loop navigation often clusters with dead clicks on the same page. Users loop back to a page, click something that looks like a link or a button but isn't, get nothing, and loop again.
When both signals are high on the same page pair, the diagnosis is usually an element that looks interactive but isn't. Something that looks like a "see pricing" link but is static text. Something that looks like an expandable FAQ row but doesn't expand.
The combination of loops plus dead clicks points to one thing: the page is promising information it isn't delivering, and users are clicking on the place where they think that information should appear.
The confusion score's element-level attribution handles this well. When 70% of loops on a page pair cluster with dead clicks on the same CSS selector, that's your target. Fix the element, not the entire page.
How to set a loop navigation alert
A reasonable threshold is a confusion score above 65 for any page that's seeing loop navigation as the dominant signal. At that level, you've got enough users looping that the pattern is statistically meaningful and not noise.
Set it tighter on revenue-critical pages. If your checkout is seeing loop navigation between cart and payment confirmation, you want to know about it fast. A threshold of 55 on checkout isn't paranoid.
The signal isn't subtle. When it fires at meaningful scale, something is wrong.