One publishing pipeline for 40 newsrooms

Kindling had grown by acquisition and inherited twelve CMS installs, four of them on versions with no upstream support left. An editor moving between titles relearned the tool each time, and a national story got pasted into six systems by hand on deadline. A platform-wide change was not slow, it was impossible — and two of the titles were one unpatched dependency away from becoming a story themselves.

One publishing core with per-title theming instead of forty codebases, so a template change lands everywhere at once. Content moved title by title through an importer that preserved every canonical URL and redirect, verified against a crawl of 1.4 million live pages before each cutover; we ran the same crawl after, and the diff was the go/no-go. Editors got a single composer with syndication built in — write once, push to any title, local edits tracked separately so a regional desk can rework a headline without fighting the national one. Delivery is edge-cached with per-title tag invalidation, so a correction purges in under two seconds instead of waiting out a TTL. We migrated the smallest title first to learn on, and the loudest one fourth, once the importer had stopped surprising us.

All forty titles were on the shared pipeline in nine months with zero redirect regressions in the post-launch crawl. Launching a new title went from eleven weeks to three days, which changed the commercial conversation more than anything technical we did. Infrastructure spend per million monthly readers fell 61%. The platform runs under a managed support plan with a fifteen-minute P1 response, and the four unsupported CMS installs are switched off.