Project · Premium appliance e-commerce
Giving 3,313 keywords a single owner each
U-Line ranked for thousands of terms and nobody could say which page owned which query. That is not a research problem, it is an ownership problem, and it shows up as your own URLs quietly competing with each other across the catalogue.
The problem
A catalogue this size accumulates keyword overlap naturally. Category pages, product pages, support pages and blog content all end up targeting adjacent terms, and without a map nobody notices until rankings stall. The site was ranking broadly and shallowly: present for a great many queries, dominant for far fewer than it should have been.
The homepage had also absorbed commercial queries that belonged on category pages, which is the most common version of this problem and the most costly, because the homepage cannot convert those visitors as well as the category built for them.
Not which keywords should we target, but which single URL is accountable for each one. Everything else follows from answering that.
What I did
- Built a keyword map assigning one owning URL per primary term, delivered as a workbook with a category map, a page priority view and a transparent removed duplicates tab.
- Ran a cannibalisation audit across the catalogue and prioritised the fixes by the search volume actually at stake, rather than by case count.
- Sized every category by keyword count and search demand, so content investment could be sequenced by revenue potential instead of by opinion.
- Built Find Your Model and Serial Number pages to capture support intent that was leaking to third party sites.
- Wrote the internal linking brief that moved commercial queries off the homepage and onto the category pages built for them.
FIGURE 1
Screenshot: the Category Map tab of the keyword workbook, client identifiers removed.
1600 x 900
FIGURE 2
Screenshot: cannibalisation report excerpt showing competing URLs for one query.
1600 x 900
Two of your own URLs swapping positions for the same term?
The cannibalisation check I run first, and what it usually turns up.
Typical reply time: within one working day