Uninventoried content will break a new architecture

What it is

After a new navigation launches, objects that never entered the inventory keep living at old addresses: expired campaign pages, landing pages a team published privately, copies in the document store, half-finished translations. People reach them from search, mail, and bookmarks, and find another world the new structure does not govern. Uninventoried content breaks the new architecture not because it is worse than the new classes, but because it proves the new structure is not the whole set. One such arrival can void the promise that “the whole site has moved.”

Gaps in the inventory and “decided to retire but left the entrance live” both create this breakage. This leaf is about missed registration itself, not the ROT verdict.

Why it happens

Architecture is a coverage claim over the reachable set. Unregistered objects falsify the claim: the same query hits the canonical page in the new class and a shadow page at the old address, and the two scents undermine each other. Old addresses still carry old labels, old templates, and old policy dates, so people cannot tell which side counts. Search engines and inbound links do not automatically honour a new sitemap; missed objects often have more ranking inertia than new pages, so the breakage pours in from outside.

If a migration defines done as “main templates swapped,” miss rate is systematically underestimated. Done has to be coverage of reachable URLs, not coverage of a visual reskin.

Studying it

Define “still alive” by reachability, not by the publishing checklist.

• Paradigms: after launch, crawl + search logs + inbound links, minus URLs the new IA claims to cover, yields the miss set; then ask participants starting on a missed page to find the canonical replacement in the new structure.
• Independent variables: inventory coverage, whether unregistered URLs redirect or show a unified notice, whether the search index was purged in sync.
• Dependent variables: arrivals on missed URLs, success at recovering the canonical page from a miss, whether people conclude there are two sites.
• Methodological note: watching only the new domain or new template misses subdomains, files on a CDN, and campaign-only hosts. Stratify the miss set by source (internal bookmarks, mail, search). Putting people on the new homepage in the lab will not detect this breakage.

Where it stops holding

A deliberate historical mirror (“2019 site archive”) is not breakage if its entrance is labelled and it is isolated from the current default lookup set. Third-party embeds (maps, payment, community) were never in the site inventory and should be treated as external, not as uninventoried. Objects published between two inventory passes are a freshness window; close it with incremental registration, not with a one-time perfect list.

Applying it

• Write the launch bar as “every reachable URL is in the new structure, 301’d to a canonical page, or an explicit historical archive.” Forbid “main pages have moved.”
• After launch, keep crawling search hits and old-template or 404 paths. Misses go on an emergency list; do not wait for the next redesign.
• Cut old search indexes, sitemaps, and automatically mailed links over to the new canonical pages. Do not only change the nav.
• Verify from the first week of post-launch search logs: queries that still resolve to old templates or unregistered paths. If those queries can still open content that conflicts with the new structure, the architecture already has a hole.

Related

• Within the group: G1.07.1 Architecture work starts with an inventory of what already exists · G1.07.2 Audits expose duplicate, stale, and orphaned content
• Adjacent: G1.11 Architectural scalability and evolution · G4.02 Deep linking · T3.03 Content freshness and maintenance
• Search terms: uninventoried content · content migration · rogue pages