Split Decisions: A Diagnostic Framework for Competing URLs on UK Sites That Grew by Accretion
Photo: Various authors., CC BY-SA 4.0, via Wikimedia Commons
Few British websites are designed. Most are accumulated. A campaign landing page is added for a seasonal push, a legacy service page survives a rebrand because nobody wanted to touch it, and a regional variant appears each time a sales lead asks for something with the city name in the H1. Nobody reconciles the job descriptions of those URLs afterwards, and that is where the trouble begins.
The important qualification is that overlap is not automatically a fault. Segmentation earns its keep when two pages serve materially different intents — a London same-day courier page and a national next-day courier page are not rivals, they are neighbours. Conflict arises only when two URLs pursue the same query with the same promise and neither is clearly subordinate.
Three tests that separate segmentation from conflict
First, intent overlap. Build a query-level export and look for terms where two or more URLs both hold meaningful impression share. Generic head terms appearing across many URLs is normal. A long-tail transactional query split evenly between two near-identical pages is not.
Second, positional instability. A single page drifting between positions three and nine across a fortnight is usually a relevance problem. Two pages trading places on the same query — one in, one out, week after week — is a selection problem, and selection problems belong to the site, not the search engine.
Third, internal support. Compare the internal links, anchor text and sitemap prominence each URL receives. Where two candidates are fed similar signal strength, the engine has no structural reason to prefer either, and the ambiguity becomes persistent rather than incidental.
Run all three before touching anything. Plenty of suspected cannibalisation turns out to be one strong page and one weak page that never had a chance; that is a pruning decision, not a consolidation one.
Ranking offenders by revenue exposure rather than impressions
The temptation is to work through a spreadsheet ordered by impression volume. Resist it. Impressions are cheap on informational pages that convert nothing, and expensive to analyse. A better priority score combines the commercial value of the query set in contention, the conversion rate attributable to each URL, the margin of the product or service concerned, and the degree of instability observed. A dedicated page or two of development to pull order data alongside Search Console exports is repaid quickly.
A worked example from a mid-market UK retailer: fourteen apparent conflicts were identified, but only three sat on queries that produced attributable revenue. Of those, one pair accounted for roughly a fifth of the category's organic contribution. The remaining eleven were retired or left alone in a single afternoon, which freed the analytical capacity to plan the three that mattered properly.
Deciding between merge, differentiate and retire
Merge is right where one URL is demonstrably stronger and the weaker page carries no unique content worth preserving. Consolidate the best elements — the clearer propositions, the better structured data, the questions the FAQ block answers — onto the survivor, redirect permanently and one-to-one, then update internal links and the XML sitemap within the same release rather than the next one.
Differentiate suits pages with genuine but muddled propositions. Sharpen each around a distinct intent, adjust titles and H1s so the distinction is legible in the results, and remove competing internal anchors that point both pages at the same phrase.
Retire applies where a page exists only for a discontinued campaign, a service no longer offered, or a region never actually served. Retirement with a relevant redirect to the nearest equivalent beats a soft 404 or a page left to decay.
Whatever you choose, resist partial fixes. A canonical tag pointing from A to B while both remain in the navigation and the sitemap sends contradictory instructions and rarely achieves what the team intended.
The ninety-day watch
Consolidation is a change to how a site describes itself, and search systems take time to accept the new description. Set expectations accordingly and monitor deliberately:
- Days 1 to 14: confirm redirects resolve in one hop, canonicals are self-referential, the sitemap reflects reality, and no internal links still point at retired URLs.
- Days 15 to 30: watch the surviving URL's position for the contended query set. Some volatility is normal; a permanent decline is not, and usually indicates that the merged page lost a content element that carried relevance.
- Days 30 to 60: compare query-level coverage before and after. The aim is not only a higher rank for the target URL but a wider set of queries it now answers alone.
- Days 60 to 90: verify commercial outcomes — sessions, add-to-basket or enquiry rate, and revenue per landing page. If a fix worked in rankings but not in revenue, the problem was never cannibalisation.
Finally, log every decision in a register that names the owner, the date and the evidence used. Accretion is not a one-off historical event; it is a habit. Sites that fix conflicts without changing the intake process — one page per intent, signed off by whoever owns search — will be running the same clinic again within eighteen months.