How we rank Indian YouTubers
CreatorVibe ranks creators each week with a five-signal weighted composite. The weights are public. The data sources are public. There is no proprietary algorithm secret — just transparent math you can audit.
Each creator gets a weekly score that is the weighted sum of five normalised signals, ranked within their (language × niche) cohort, with movement arrows showing how they shifted versus last week.
The five signals
Each signal is calculated from raw YouTube Data API metrics, then normalised within the cohort (creators in the same language and niche). Normalisation means a 5% engagement rate scores differently against the Telugu Tech cohort versus the Tamil Finance cohort — fair comparisons within peer groups, not absolute numbers.
| Signal | Weight | What it measures |
|---|---|---|
| Audience love | 25% | 0–100 score from sampled comment sentiment (Gemini 2.5 Flash) across each creator's recent videos. Counts positivity, comment depth, cultural-affection markers (regional address terms, possessive "mana"/"ours" language, emoji), and niche-appropriate engagement. Falls back to an engagement-rate proxy when the comment sample is too small. |
| Engagement rate | 25% | Average (likes + comments) ÷ views across the creator's videos from the last 30 days. Penalises hate-watching and bot-inflated views. |
| Subscriber growth | 25% | Week-over-week % change in subscriber count. Rewards momentum, not size. |
| View velocity | 15% | Week-over-week % change in total channel views. Measures whether content is breaking out. |
| Consistency | 10% | Upload rhythm over recent videos — measured from publish dates, blending a healthy cadence (≈0.5–4 uploads/week) with regularity (even gaps beat erratic bursts). Steady uploaders score highest regardless of whether they posted this exact week. |
Weights add to 100%. Default weights apply across all niches; future versions may calibrate per-niche weights once we have enough data per cohort.
Where the data comes from
- YouTube Data API v3 — official, public, rate-limited to 10,000 quota units per day. We use
channels.list,playlistItems.list, andvideos.list. Neversearch.list(100 quota units, too expensive at scale). - Derived signals (Audience Love, Consistency, Engagement) — computed by CreatorVibe from successive daily snapshots and recent-video stats. The math is public; the data is yours.
- Comment sentiment (active) — we pull a sample of comments per video via the public YouTube Data API and use Gemini 2.5 Flash to classify positivity, emotional depth, and cultural-affection markers. This now drives the Audience Love score directly, falling back to an engagement-rate proxy only when a creator has too few sampled comments.
- What we do not use — Watch time, audience demographics, revenue, and retention curves. These are owner-only via the YouTube Analytics API and can only be unlocked once a creator claims their CreatorVibe profile and grants Analytics access.
How often it updates
Two pipelines run on separate cadences:
Daily — every day at 08:00 IST (02:30 UTC):
- Pull fresh channel + video stats from YouTube for every tracked creator
- Write daily snapshots to our database (rolling 30-day window for ToS compliance)
- Prune raw snapshots older than 30 days; derived rollups are kept permanently
Weekly — every Monday at 09:30 IST (04:00 UTC):
- Compute weekly rollups (growth %, engagement over 30 days, consistency over recent uploads)
- Run comment-sentiment analysis to set each creator's Audience Love score
- Re-rank the leaderboard, hidden gems, rising stars, and audience love boards
We rank weekly because that is the cadence creators and audiences actually think in — and because engagement signals are measured over a rolling window, not a single day. The leaderboard URL /te/tech reflects the latest weekly run within ~60 minutes of completion.
What the movement arrows mean
Each creator's rank is compared against last week's rank for the same board:
- ▲ 4 — climbed 4 positions this week
- ▼ 2 — dropped 2 positions this week
- NEW — did not appear in last week's top-N
- — — same rank as last week
Note: until CreatorVibe has been running for at least two consecutive weeks, every entry shows NEW because there is no previous ranking to compare against.
Cohorts: why rankings are within (language × niche)
A 50K-subscriber Telugu cooking creator and a 5M-subscriber Hindi tech creator are not competing for the same audience or the same advertisers. Ranking them on the same board is meaningless. Every CreatorVibe ranking is scoped to a cohort — a single (language, niche) pair like Telugu Tech, Tamil Finance, or Hindi Gaming. Engagement-rate normalisation happens within that cohort, so a 4% engagement rate scores differently against a tech cohort (high baseline) than a comedy cohort (lower baseline).
Hidden gems and rising stars
Beyond the main leaderboard, two boards surface creators the algorithm buries:
- Hidden Gems — creators with under 50,000 subscribers whose engagement rate is at least 3× the cohort median. They have rabid, small audiences that YouTube's discovery prefers not to amplify.
- Rising Stars — creators with subscriber growth above 10% week-over-week and consistent upload cadence. These are the breakout signals before they break out.
Why this is hard to game
The ranking weights ratios over absolutes. Specifically:
- Buying subscribers without engagement lowers your engagement rate (25% of score) and audience love (25% of score) — net negative.
- Bot views dilute your engagement rate similarly. View-velocity alone (15% weight) is not enough to climb.
- Burst-uploading hurts consistency (10% weight) — the score rewards even gaps, not erratic bursts. Steady cadence wins.
- Half the weight now sits on genuine audience signals (love + engagement), which are far harder to fake than subscriber or view counts.
- We compare to cohort, not absolute. Inflating numbers against a strict peer comparison is harder than inflating against an absolute threshold.
Audience Love runs on real comment-sentiment with spam filtering, so bot comments are detectable and don't inflate the score.
Claiming your profile
If you are a creator on the leaderboard and want to claim your profile, see your private score breakdown, or correct your data, head to the FAQ (specifically the "Can I claim my channel?" section). The claim flow is rolling out — for now, DM us via the contact link in the FAQ and we will get you sorted.
Versioning and changes
Ranking weights are stored as versioned configuration in our database. When we adjust them, we do it as a new version row, not an edit — so we can audit and roll back. The current version is 2026-05-29.v2.
What changed in v2 (2026-05-29):
- Retired the discovery velocity signal — a log-scale view proxy that added little the other signals didn't already capture.
- Reweighted toward genuine audience signals: Audience Love 15% → 25%, Engagement 20% → 25%; Subscriber growth 30% → 25% and View velocity 20% → 15%.
- Engagement now measured over a rolling 30-day window instead of the current week, so creators aren't penalised for a quiet week.
- Consistency now measured from upload rhythm (cadence + regularity over recent videos) instead of a single week's upload count.