Discovery & Matching

Search & filtering (v1 — deterministic, not ML-driven)

The v1 discovery experience should be a well-built faceted search, not an AI black box — predictable, debuggable, and buildable without a data-science team. Filters:

• Category/niche (multi-select taxonomy — see 04-features-admin.md for taxonomy management).
• Platform (Instagram / YouTube / TikTok / LinkedIn / X / Podcast).
• Follower/subscriber range (nano <10K, micro 10K–50K, mid 50K–500K, macro 500K+ — configurable bands).
• Engagement rate range.
• Price band (by the creator's rate card).
• Location/city, language.
• Verification status (identity verified / audience verified / both).
• Availability (open to gigs now).
• Rating threshold (4+ stars, etc.).

Sort options: relevance (default — see ranking below), price low-to-high/high-to-low, engagement rate, follower count, most reviewed, recently active.

Ranking signals (default "relevance" sort)

A weighted composite, computed server-side, roughly:

1. Query/filter match strength (how well the profile matches the active filters/keywords).
2. Reputation (rating, completed gigs, on-time rate).
3. Responsiveness (average response time to pitches/messages — a strong predictor of good hirer experience).
4. Recency of activity (active-in-last-30-days ranks above dormant profiles — keeps stale/abandoned profiles from cluttering results).
5. Verification tier (verified profiles get a ranking boost — incentivizes going through verification, improves overall search quality).
6. Paid featured placement — a bounded, clearly-labeled boost slot (see 06-business-model.md), never allowed to override relevance entirely (e.g., cap featured placements to 1-in-5 result slots, always labeled "Featured").

Gig-side discovery (for creators)

Mirror structure: creators filter/browse open gigs by category, budget range, deliverable type, deadline, and get gig alerts — saved-search notifications ("notify me when a fashion-niche Instagram gig with budget >₹15K is posted") — this is a high-value, relatively simple feature that directly drives creator engagement/retention.

Matching assistance (Phase 2+)

Once there's enough transaction data to train on:

• "Creators like this one" recommendations on a profile page.
• Brief-to-creator matching: hirer describes their campaign in a few sentences, system suggests a ranked shortlist — genuinely valuable but requires either (a) real ML investment or (b) a well-tuned heuristic/keyword-matching system as a credible v1.5 stand-in before investing in ML. Do not overbuild this pre-launch; a good faceted search covers the real v1 need.
• Lookalike audience matching: "this creator's audience overlaps with your existing customer base" — powerful but dependent on deeper platform-API data access that's a longer-term integration effort (see 10-technical-architecture.md).

Explicit non-goal for v1

Do not attempt a fully automated "AI picks your influencer for you" experience at launch. It requires data the platform won't have yet (no completed-campaign history to learn from), risks eroding trust if recommendations are bad early on, and distracts from getting core marketplace liquidity (enough real creators, enough real gigs) working first. Filtered search + gig alerts is the correct, honest v1 scope.