Content Discovery Feature Template

Search serves explicit intent: the user types a query and expects matching results. Discovery serves latent intent: the user does not have a specific query but is open to relevant suggestions. Most products need both. Search handles "I know what I want." Discovery handles "Show me what I should explore next." For a full search system design, use the [Search Specification Template](/templates/search-spec-template).

Use content-based recommendations when you have strong metadata (tags, categories, descriptions) but limited user interaction data. Use collaborative filtering when you have abundant user behavior data and want to surface non-obvious connections. Most production systems use a hybrid: content-based for cold-start, collaborative for established users.

Set explicit diversity guardrails: cap per-category representation, reserve slots for exploratory content, and inject items from adjacent categories the user has not yet engaged with. Monitor diversity metrics (Shannon entropy of category distribution) alongside engagement metrics.

Start with one. Pick the highest-traffic surface (usually the homepage) and build a solid recommendation pipeline for it. Once the infrastructure is stable and you have measurement in place, expand to secondary surfaces (post-completion, email digests, sidebar widgets). Trying to launch five surfaces simultaneously splits your team's focus and delays learning.

The gold standard is "content breadth": are users engaging with more types and topics of content than they were before? Secondary metrics include long-tail reach (is your back catalog getting more views?), time-to-first-engagement, and user satisfaction surveys. Watch for vanity traps where CTR goes up but actual value delivered stays flat. ---