Media & Publishing
A publisher's core loop is headlines, authors, and scheduling - and headline testing is one of the oldest, highest-return uses of A/B testing there is. This walkthrough sets up an Articles collection and a headline experiment.
Modeling articles
| Field | Type | Notes |
|---|---|---|
| headline | text | The single highest-impact field on the entry |
| body | richtext | Article body |
| author | relation | Points at an Authors entry |
| heroImage | media | |
| section | enumeration | e.g. news / opinion / culture |
| publishAt | date | Scheduled publish time |
| breaking | boolean | Pins the article to the top of its section |
Scheduling and sections
A section front fetches the latest published articles for that section, newest first:
?where[section]=news&sort=publishAt:desc&limit=20Testing headlines
Duplicate an article entry, change only the headline field, and run the two as an experiment weighted 50/50. Your front-end or CMS-facing edge function requests the article with ?experiment=1 so each pageview gets a headline variant, and fires a conversion on a click-through to the full article:
POST /public/v2/projects/:uuid/experiments/variants/:variantId/convertOnce a clear winner emerges, declare it from the Experiment tab - the underlying entry that everything else (the CMS search index, internal links, the author's byline page) already points at does not change; you are only picking which headline variant the experiment resolves to.
Breaking news