Abigail Sutherland - Organizing a Confluence hoard, or, does this page spark joy?

A sprawling Confluence setup—33 spaces and more than 20,000 pages—was turning into a daily irritant because people couldn’t find what they needed and teams kept duplicating or letting content rot. Abby Sutherland, TomTom’s Nav information architect, tackled the problem by redesigning the information architecture and then using usage data to decide what deserved to stay, move, or be archived—turning “does this page spark joy?” into a practical, measurable cleanup workflow.

The effort started with a commitment to tidying up, but the real leverage came from getting user buy-in. Once Sutherland took ownership of the Confluence mess, feedback poured in: people surfaced missing context, outdated pages, and pain points that could be used to shape the redesign. From there came a “future site” mindset—either building a greenfield space or restructuring an existing one—paired with a key diagnostic: look for “desire paths.” If users are creating workarounds (like walking off an official route), it signals an unmet need. In Confluence terms, that means fixing the underlying reasons people can’t reach content—often by improving search scope, simplifying browse paths, and surfacing deep content at the top.

Sutherland’s strategy centers on user-first design, but with a hard-nosed view of constraints. Confluence search doesn’t behave like Google, and learned helplessness sets in quickly when people can’t find old material. Tool limitations also shape behavior: Confluence requires unique page names within a space, which makes standardization tricky and can create menu clutter when teams namespace pages. Even tagging and workflow-based review are difficult, so content quality depends heavily on whether teams feel responsible.

To manage those constraints, she structured the hierarchy using a “golden line” between centralized control and team autonomy. The top layer is tightly controlled through a small set of category areas and standardized team landing pages. Those landing pages include a search box and page tree that search only within the team’s space—intentionally narrowing scope to improve signal-to-noise in Confluence search. Below that layer, teams own their internal page structures and can innovate without breaking the overall navigation model.

The cleanup itself relied on a “joy detector.” Instead of trying to manually judge 20,000 pages, Sutherland used Confluence’s API and internal reporting to pull metrics such as page view behavior and last-view timing. After testing candidate metrics, she built a weighted “view count” formula (average daily views divided by days since last viewed, with safeguards against division by zero). She then generated heat maps—color-coded page trees—to highlight which pages were actively used versus likely deadwood. Teams reviewed those maps, migrated what mattered, and archived what didn’t.

By the end of the work so far, five spaces were deleted outright, 13 spaces had their good content extracted and the rest archived, and three spaces were reclassified after organizational changes. The remaining work focuses on the last spaces, plus building a maintenance checklist and schedule so the system doesn’t drift back into chaos. The new consolidated space already exceeds 10,000 pages, but it’s structured to mitigate large-space weaknesses, with average content age and last-view metrics roughly halved compared with the old setup—evidence that the architecture and the data-driven triage are improving discoverability and freshness.