TikTok Influencer Audit — Engagement & Sponsored Rate Estimates

TikTok Influencer Audit turns sampled public performance into quality scores, risk signals and sponsored-video planning rates—not definitive bot/fraud verdicts, an audience census, creator quotes or conversion forecasts.

Modes and actual collection behavior

Mode	Behavior
collectAndAnalyze (default)	Requires creator handles/profile URLs; starts one grouped Creator Videos child run and merges any supplied video datasets.
analysisOnly	Requires videoDatasetIds; reuses datasets and starts no scraping Actor. It still attempts live public profile enrichment via a third-party HTTP endpoint. It is not an offline mode.

There is no mock mode. Both modes read supplied profile/comment/audience datasets. Analysis-only infers usernames if creators is omitted. Live resolved profiles override supplied profiles, including unavailable fields.

depth is a collection preset, not a gate on supplied evidence:

• quick: no live comment child run.
• standard/deep: collect mode adds one grouped comment child after videos/profiles. URLs are selected by source-video view rank, not necessarily the final pinned/window-filtered audit sample; at most 100 URLs, replies disabled.
• Audience is never collected here. Supplied audience/comments are analyzed even in quick; Deep requests no TikTok cookie.

Child calls precede scoring/saving, with separate event budgets and fee/usage costs even if no report is scored. Parent budget is not an all-inclusive ceiling.

Quick start

Illustrative live input:

{
  "mode": "collectAndAnalyze",
  "creators": ["mrbeast"],
  "depth": "quick",
  "maxVideos": 5
}

To reuse existing data without a video/comment child run:

{
  "mode": "analysisOnly",
  "videoDatasetIds": ["YOUR_VIDEO_DATASET_ID"],
  "depth": "quick",
  "maxVideos": 30
}

Save either input as input.json for API use:

curl -sS -X POST \
  "https://api.apify.com/v2/acts/yICqJ7Z5GgJeNWuAB/runs" \
  -H "Authorization: Bearer ${APIFY_TOKEN}" \
  -H "Content-Type: application/json" \
  --data-binary @input.json

Inspect completion/status/logs, dataset and OUTPUT. No TikTok password/cookie/API key is requested; Apify authentication and dataset access are required.

Input reference

Field	Default	Supported options / range
mode	collectAndAnalyze	collectAndAnalyze, analysisOnly.
creators	none	Usernames, @handles, profile URLs; public schema accepts 1–20 entries. Analysis-only may infer creators instead.
videoDatasetIds	[]	Required for analysis-only; also merged with live collected videos in collect mode.
profileDatasetIds	[]	Optional compatible public profiles, such as Finder output.
commentDatasetIds	[]	Optional comments mapped by creatorUsername or a known videoId.
audienceDatasetIds	[]	Optional profiles mapped by sourceUsername or creatorUsername.
depth	quick	quick, standard, deep; see actual collection behavior above.
maxVideos	30	5–100 selected videos per creator.
analysisWindowDays	180	30–730; fewer than five recent valid videos triggers selection from all valid videos, then maxVideos.
maxCommentVideos	2	1–5 representative source videos per creator for live comment collection.
commentsPerVideo	50	10–300; returned quantity is not guaranteed.
brandKeywords	[]	Caption substring fit; omitted means unavailable, not guessed.
market / niche	GLOBAL / general	Recorded rate assumptions, no market/niche price multiplier.
cpmLowUsd / cpmRecommendedUsd / cpmHighUsd	5 / 10 / 15	USD per 1,000 views; schema 0.1–500; require low ≤ recommended ≤ high.
minimumCreatorFeeUsd	50	0–100000; applied before quality adjustment and production fee, not a guaranteed final quote floor.
productionFeeUsd	0	0–100000; fixed addition to each base estimate.
usageRightsDays / sparkAdsDays	0 / 0	0–730; modelled add-ons only.
exclusivityDays	0	0–365; modelled add-on only.
maxConcurrency	2	1–5; profile concurrency and forwarded to grouped child runs.
requestTimeoutSecs	20	10–60 seconds per live profile HTTP request.

Each supplied dataset is read once, up to 10,000 rows, without pagination. Videos need username, ID/URL and positive views; compatible keys include authorUsername/author.uniqueId, id/videoId, playCount/viewCount, createTimeUnix/createTimeISO, diggCount, commentCount, shareCount, collectCount. Timestamps accept seconds, milliseconds or ISO dates. Unknown dates remain eligible; missing engagement counts become zero and can lower scores.

Scores, rates and output

Scoring uses deduplicated, date-sorted, positive-view, non-pinned videos; at least five are required. The 0–100 model weights available reach 20, engagement 20, consistency 15, activity 10, public audience 15, brand fit 10 and readiness 10, then applies risk penalties. Missing components leave the denominator. Labels: STRONG 80+, GOOD 65+, WATCH 50+, HIGH_RISK_SIGNALS below 50, or INSUFFICIENT_DATA. Confidence is completeness, not accuracy probability.

Comment patterns use text repetition, author diversity and short/emoji-only text. Audience patterns use sample uniqueness, posting/privacy flags and region fields—not bot/fraud proof or protected demographics. Caption brand/sponsorship markers are heuristic; identity and contact delivery are unverified.

Rate estimates use view quartiles and CPM, with USD rounded by amount:

low = max(floor, P25 views / 1000 × low CPM) + production
recommended = max(floor, median views / 1000 × central CPM) × Q + production
high = max(floor, P75 views / 1000 × high CPM) × max(1, Q) + production

Q is a score-based factor bounded to 0.75–1.2. High is raised to at least recommended. Rights add-ons are summed, then applied to the base including production: usage +25% per started 30 days (cap 100%), Spark Ads +20% (cap 60%), exclusivity +20% (cap 80%). These are transparent model assumptions, not market-verified contract prices. A recommended estimate can fall below the input floor after quality adjustment.

One row is one report: creator, audit (components/strengths/risks/gaps), metrics, pricing, source (selected/dataset IDs, child runs, analysis time). Engagement rates are fractions: 0.08 means 8%. OUTPUT holds computed reports, versions (quality-v1.0.0, view-cpm-2026.08-v1) and billing. No scoreable reports means failure after possible diagnostic writes.

Current event pricing and timing

Event	FREE	BRONZE–DIAMOND
apify-actor-start	$0.0022/event	$0.0022/event
video-analysis	$0.00625/video	$0.00125/video
comment-audit-start	$0.0625	$0.0125
comment-analysis	$0.00625/comment	$0.00125/comment
audience-profile-analysis	$0.00005/profile	$0.00001/profile

Startup is one event per GB of allocated memory, minimum one event per run, including empty/failed runs. Platform compute, transfer and storage are separately billed to the running user; child Actor fees/usage are additional. The configured minimum parent event-budget ceiling is $0.04, not a minimum bill.

Custom analysis events are not per-report fees. Scored reports count selected videos and object-row comment/audience samples, including supplied data at any depth. Setup is once per completed live comment child plus at least one scored report; dataset-only reuse has no setup event.

Maximum live-analysis budget is checked before child calls, then computed-event capacity before saving. Save precedes custom charges: a later charge error can leave rows in a failed run. Insufficient-data reports have no analysis events, but startup/child/usage costs remain.

The developer-set Free cap is 50 report rows globally, not 50 analyzed samples or an analysis-cost ceiling. Custom counts/OUTPUT use computed reports before truncation. Paid plans have no developer-set 50-row cap. OUTPUT.billing.listEventChargeUsd uses base prices/one planned start, not a tier/memory-accurate invoice. Use charged-event counts and the Pricing tab.

FAQ and limitations

Is analysis-only completely offline? No: it starts no scraper but still reads Apify datasets and attempts live profile enrichment.

Does Deep collect followers? No. Audience analysis requires a supplied compatible sample; it is not a census and may be biased or stale.

Can the score identify bots or guarantee a sponsorship rate? No. Signals and rates are estimates for human review, not definitive judgments, guaranteed quotes or predicted sales.

Source/dataset limits, child Free caps and budgets can reduce samples or fail collection. No ownership, full-history, reach or brand-safety guarantee. Respect platform terms, privacy and lawful data use.

Complementary Actors

• TikTok Influencer Finder — discover a shortlist and public profile/contact data.
• TikTok Creator Videos — prepare reusable public-video datasets.
• TikTok Comments Scraper — prepare reusable comment samples.