Creator profiles — engagement analytics, audience data and fake-follower checks

What a dossier contains

• Identity: handle, display name, bio, verified status, cross-platform links (Instagram, YouTube, X, Douyin).
• Performance: followers, average / median / max views, average likes, engagement rate, view-to-follower ratio — all computed from the creator's scraped videos.
• Behavior: account age, posting cadence (videos per week), first/last video dates.
• Audience: inferred country and content category, plus the sampled audience snapshot below.
• Contacts: email, website, Instagram, YouTube, Linktree — harvested from bio links.
• Commercials: tier (nano/micro/macro/mega), estimated price range, authenticity score.

Dossiers live at the KOLs page in the app (GET /api/kols/{username} on the API) and every creator also has a public, shareable dossier at kolens.ai/k/<username> — handy for dropping into a client Slack.

Why median views matter more than followers

Follower count is the easiest number to inflate and the worst predictor of campaign results. KOLens leads with view metrics computed from actual recent videos: median views resists the one-viral-video distortion, and the view-to-follower ratio exposes accounts whose audience stopped watching. A 100k-follower account averaging 2k views is a weaker buy than a 30k account averaging 25k — the dossier makes that visible in seconds.

Authenticity score (fake-follower check)

The authenticity score (0–100, higher = more authentic) combines 8 heuristic signals computed from data KOLens already has — no extra scraping, no extra cost:

• Follower-growth spikes (sudden 5× jumps are suspicious; steady 2–5% weekly is healthy)
• Engagement-rate plausibility versus the account's tier band
• Comment-to-like ratio (bot-boosted accounts have likes but no conversation)
• View dispersion across videos and median/mean view distribution (purchased views flatten the curve)
• Audience geo consistency and demographic plausibility
• Posting-cadence regularity (humans are irregular; farms are clockwork or burst-silent)

Each score ships with a confidence tier (insufficient_data / low / medium / high) based on how many signals could fire — a new account with 3 posts gets honesty about uncertainty, not a fake precise number.

Audience snapshot

The audience snapshot (GET /api/kols/{username}/audience) samples the creator's followers and reports top countries, languages, follower-tier distribution, top interests/hashtags, active-follower share, and a confidence interval on each estimate. Use it to verify the question that decides cross-border budgets: is this audience actually in my target market?

Price estimate

KOLens estimates a fair per-video sponsorship range with a CPM-on-views model: expected views (capped at follower count) × a tier CPM (nano $30 / micro $25 / macro $18 / mega $15 per 1,000 views) × an engagement multiplier (≥8% ER → 1.25×, ≥4% → 1.1×, <2% → 0.85×), widened to a min–max band. Example: 50k followers, 5% ER, 10k average views → $206–$385 per video. Use it as a negotiation anchor against the creator's quote.

How to vet a creator in 3 minutes

1. Open the dossier and check median views against follower count — the view-to-follower ratio is the fastest health check.
2. Check the authenticity score and its confidence; below 30 with medium+ confidence is a hard pass.
3. Open the audience snapshot — confirm the top country matches your target market.
4. Compare the price estimate with your budget; the range plus the engagement multiplier tells you whether a premium is justified.
5. Optionally pull the activity summary (free cached; force refresh 1 credit) — an AI narrative of the last 7/30/90 days including sponsored-content frequency.

Advanced usage

• Lookalike expansion: find_similar_creators (MCP) returns 20–50 similar accounts ranked by co-mentions, category, tier and shared hashtags — grow a niche from one proven creator.
• Premium AI deep-dive: the Deep Overview (2 credits) writes a holistic brand-fit read; background research (1 credit, free on cache hit) adds web-grounded context beyond the platform data.
• LLM-friendly rows: GET /api/kols?llm_friendly=true adds a pre-digested summary, risk_flags and next_actions per creator — built for agent consumption.
• Free public audit: the /tiktok-audit tool runs a lightweight fake-follower check on any handle without an account.