Key pages four clicks deep: checking click depth in JMX

The shop has 8,400 pages. The category pages that carry the revenue are reached from the home page only via the overview, a filter page and two rounds of pagination. Nobody noticed: everything is linked, everything answers 200. It is expensive all the same, because the deeper a page sits in the click structure, the less internal link equity reaches it and the less often it is re-fetched — on a site of this size, click depth is one of the few strong levers on crawl budget that can be moved without a line of new content.

How to go about it in JMX

1. Start the crawl without a tight limit. Under Settings → Crawl sit the maximum depth and the URL count. Set both high enough. A run that ends at depth 4 reports no page at depth 6 — it does not know it exists.
2. Read the distribution by depth on the Dashboard. A click on a bar opens the Results, filtered to that click depth. Next to the number of hits sits the label “Depth 3” — the only hint that the list is trimmed.
3. Set the threshold. The rule “page sits too deep in the click structure” reports every page deeper than the setting — the default is 5 clicks, so reporting starts at depth 6. The interface has no slider for it, though: under Settings → Rules, rules can only be switched on and off and have their severity changed, and the Thresholds tab does not carry click depth. Whoever changes it does so in the stored settings file; after that, Refresh in the Issues view is enough, a second crawl is not needed.
4. Show the link score. In the result tables the Link score column can be switched on, the internal PageRank from 0 to 100. Depth says how far away a page sits, the link score says where it is linked from: three links from the home page carry more than thirty from a forgotten tag archive.
5. Look at the crawl tree. It draws the click depth per page. Colour by depth and, if Search Console data is loaded, set the node size to impressions. That way you see in one picture where the demand lives and how deep it sits.
6. Order in the Tasks tab. Impact in points, estimated effort and leverage — points per person-day — sit there together with the classification into Do now, Schedule, Alongside and Defer. The work falls per affected page: 300 pages that sit too deep are a project, not a quick fix.

What to watch out for

Click depth is not path depth. An address with five directory levels, linked straight from the home page, has depth 1; a short /offer/ can sit at depth 6. Measurement starts at the start address: begin the crawl at /shop/ and you get a distribution that looks two levels too good.

The depth comes out of the links in the HTML as served. What a script only supplies afterwards — a mega menu, a “load more” button — the crawl does not follow, and most AI answer crawlers do not execute JavaScript either. JavaScript rendering does show which links only come into being in the browser, but it does not recompute click depth — and it is no substitute for navigation that is never served from the server.

The finding comes as a notice, not a warning: a deep page is not a mistake but a decision that can be wrong — filter the Issues view to severity “Warning” and you never see the row. One special case is pagination: every page of it creates a level, and shortening the pagination only moves the problem; what helps is a topic overview that points directly at the detail pages.

If the distribution is flatter on the next run and the pages still get no clicks, depth was not the problem — and you know that after an afternoon instead of after a quarter.