Recommendations

To see which entries are most like a given one, choose Recommendations → Similar Content in the sidebar and pick the reference entry from the menu at the top. Content lists up to eight similar entries, each with its type, status and a score, highest first.

When the project's entries have been indexed for semantic search, the ranking uses their meaning, with your boosts and exclusions applied, and Method shows Semantic. Otherwise Content falls back to comparing the words in each entry's title, excerpt and body, Method shows Keywords, and a note says why semantic results aren't available. The figures across the top also show how many Candidates were compared and the Top score.

The Recommendations screens rank your project's entries so you can review them inside Content. Two kinds of editorial rule adjust those rankings:

• A boost multiplies an entry's score by its weight, so it ranks higher.
• An exclusion removes the entry from rankings.

You add them under Personalization → Boosts & Exclusions or under Recommendations → Business Rules. Both screens edit the same set of rules; Business Rules also records a reason for each one. Paused rules don't apply. Screens that can't measure real traffic label their scores as estimates.

Recommendations → Content-based Ranking shows how the entries most like a reference entry rank once your boosts and exclusions apply. Pick the reference entry from the menu at the top. Content lists up to 10 entries with their final score, and marks the ones a boost lifted as Boosted. Excluded entries don't appear.

The figures across the top count the entries Ranked, how many are Boosted, and the active exclusion rules (Excluded). If every candidate is excluded, or the project has only one entry, the list shows "Nothing ranked".

Recommendations → Trending Content lists the 10 entries updated most recently, with a trend score that falls as an entry ages. It doesn't measure views: the screen labels every score and age as an estimate, and Measured traffic shows None. Use it as a quick look at what's fresh, not at what's popular.

Recommendations → Context-aware Ranking lets you try the slot decision rules without any traffic. It uses the same matching as real slot decisions.

1. In Variants (JSON), list the variants to test, each with an id, entryId, priority, weight, status and rules. The box starts with an example.
2. In Context (JSON), describe the request, like { "device": "mobile", "locale": "en" }.
3. Choose Run.

Outcome shows whether each variant matches the context and how many clauses it has, then the winning entry and the reason. Nothing is saved. If either box isn't valid JSON, Content says "One of the JSON blocks is invalid." Personalization explains how slots decide.

Business rules are boosts and exclusions with a recorded reason, such as a sponsorship or a compliance block. To add one, choose Recommendations → Business Rules, then Create rule:

• Entry: required.
• Rule type: Boost (the default) or Exclude.
• Weight: the boost multiplier, 2 to start. Exclusions don't use it.
• Business reason: required, like Sponsor Q4.

Choose Create rule. New rules are Active. Each rule's ⋯ menu has Pause or Activate and Delete, which removes it straight away, without asking. The figures across the top count active Business rules, Sponsor boosts and Compliance excludes. Search entries and reasons. Changing rules needs the Manage settings permission, which the Owner and Admin roles have.