Audience Persona is live — sampled-follower analytics for any TikTok creator

Why a sampled snapshot vs full enumeration

Two reasons. First, it's the only thing that's affordable: a 1M-follower creator fully enumerated would cost ~$1000+ in Apify spend per refresh. A 500-sample is $1.20 to us and 30 credits to the user. Second, it's all you actually need: a proportional 500-sample is statistically equivalent to the full enumeration for the questions a brand actually asks ("are these followers in our shipping markets?", "what % is active?", "what else do they engage with?").

We make the statistical reality the loudest thing in the UI — the footer reads "Sampled 500 followers · 95% CI · ±4.4% error" on every snapshot. Marketers can quote the headline number with the caveat baked in.

What's in the panel

1. 1

   Country mix — top 8 + Other.

   Bar chart with each country's percentage of the sample. Helps answer "is this audience in markets we can ship to?" and "should we localise the creative for region X?".
2. 2

   Language — top 4.

   Same shape; different axis. Catches creators whose follower base has drifted away from the language of their captions (very common for global tech / sports / fitness niches).
3. 3

   Activity % — last 30 days.

   Big number: what fraction of sampled followers posted at least one video in the last 30 days. The brand-vetting "are these real, active accounts?" check.
4. 4

   Follower tier distribution.

   Each sampled follower bucketed by THEIR own follower count (nano / micro / mid / macro / mega). Reveals whether the creator is followed mostly by other consumers (nano-heavy) or by other creators (micro+).
5. 5

   Top niches — from the active cohort only.

   Hashtag pills weighted by frequency in the active-cohort's recent videos. The active-only restriction matters: passive consumers' "favourite hashtag" data is mostly noise from videos they liked once. Restricting to creators who post in the last 30d filters that out and gives signal.

Drift detection (AP-4)

Enable audience_auto_refresh on a watched KOL and KOLens re-samples every 7 days (configurable per row). After each refresh, the new snapshot is compared against the previous; if any country's share moved by ≥10 percentage points, an audience_drift alert fires.

This is the early-warning signal for "the audience that engaged with our last campaign isn't the same audience now." Common causes:

• Viral hit shifts the demographic — a video that broke through to a different language / country dominates the recent-followers pool.
• Country-restricted shadowban — sudden drop in one region's share without a corresponding rise elsewhere.
• Bot wave — rapid follower-count growth concentrated in one country with low activity %; usually a flag to slow down sponsorship spend until it normalises.

The drift alert pushes through the same email + webhook delivery channel as rising_kol / trending_video alerts.

Pricing

Pre-charged at request time per the standard tier table:

• 200-sample → 15 credits · ±7% error · ~1-2 min run
• 500-sample → 30 credits · ±4.4% error · ~2-5 min run
• 2000-sample → 100 credits · ±2% error · ~10-15 min run

Apify failures auto-refund. Auto-refresh on watchlist rows charges the user's credits at the same rate per refresh (snapshot quality + audit trail are valuable enough that we don't subsidise the recurring cost the way we do for FB Ad Spy watch refresh).

Honest limits — read these before you trust a number

1. Statistical sample, not full enumeration. Show CI + error margin on every panel; we do, you should reproduce when you quote.
2. No age, no inferred gender in v1. Don't approximate them from "looks like a beauty creator follows" — that's confounded.
3. Comparable to but not identical to TikTok Studio. Studio has full panel coverage + ML inference; we have a sample of public profiles.
4. Snapshot is a moment-in-time. Use the drift chart (when there are ≥2 snapshots) before making campaign decisions on the latest single one.

Next step

Try it on a creator you already know — how does KOLens' country mix compare to the screenshot they showed you from TikTok Studio? When the numbers line up, that's calibration you can carry into vetting unknown creators.