How It Works

V2 pricing model — combines real X data with AI analysis, time-decay weighted impressions, and a scarcity factor for transparent KOL tweet pricing.

Analysis Pipeline

Fetch User Profile

We call the X API v2 to get the KOL's profile: followers, following, listed_count, bio, and account age.

Fetch 30 Recent Tweets

We pull the last 30 original tweets (excluding retweets and replies) with full engagement metrics: impressions, likes, replies, retweets, quotes, bookmarks, and entities.

Trim Top/Bottom 3

We remove the 3 tweets with the highest impressions and the 3 with the lowest. This eliminates extreme outliers, leaving 24 tweets.

IQR Anomaly Detection

We apply IQR 1.5x filtering on the remaining tweets to detect and remove statistically anomalous impressions (e.g., bought engagement).

Claude AI Analysis

Claude AI analyzes the bio and tweets to determine: Domain + SubDomain (10 subcategories), Credibility score (0-100), Relevance score (0-100), and Identity tag.

4-Dimension Scoring

We calculate a composite Overall Score from 4 weighted dimensions: Influence Depth (20%), Follower Quality (40%), Content Stability (25%), and Engagement Quality (15%).

V2 Pricing

CPM = $5 + (Score/100) × $55 (range $5~$60). Price = CPM × (WeightedImp/1000) × Domain × Credibility × Relevance × Identity × Scarcity. Range is ±20%.

V2 Pricing Formula

Overall Score = ID×20% + FQ×40% + CS×25% + EQ×15%

CPM = $5 + (Overall Score / 100) × $55

Weighted Impressions = Σ(impressions × time_weight) / Σ(time_weight)

Modifiers = Domain × Credibility × Relevance × Identity × Scarcity

Price = CPM × (Weighted Imp / 1000) × Modifiers

Range = Price × 0.8 ~ Price × 1.2

Time Decay Weights

Recent tweets are weighted more heavily — older viral tweets don't inflate today's reach estimate.

Impression Time Decay

≤ 7 days ago1.0x

8 – 14 days ago0.8x

15 – 30 days ago0.6x

> 30 days ago0.4x

Weighted Engagement

Different interaction types carry different transmission value. Used for Follower Quality (ER) and Engagement Quality scoring.

Engagement Value Weights

Quote (×4)Highest — carries a comment, max spread

Reply (×3)Real discussion signal

Retweet (×2)Second-degree reach

Bookmark (×2)Long-tail value

Like (×1)Baseline

4 Scoring Dimensions

1. Influence Depth (20%)

Composite of 2 sub-items: Follower Scale (60%) + Listed Ratio (40%)

Sub-item 1 — Follower Scale (60%)

> 100K100

50K – 100K80

20K – 50K60

10K – 20K40

< 10K20

Sub-item 2 — Listed Ratio (40%) — listed_count / followers × 1000

> 5100

3 – 575

1 – 350

0.5 – 125

< 0.510

2. Follower Quality — Weighted ER% (40%)

ER = Average Weighted Engagement / Followers × 100%. Weighted engagement = likes×1 + replies×3 + retweets×2 + quotes×4 + bookmarks×2.

> 2%100

1% – 2%75

0.5% – 1%50

0.3% – 0.5%38

0.1% – 0.3%28

0.05% – 0.1%20

< 0.05%10

3. Content Stability — Combined CV (25%)

Combined CV = 0.4 × posting interval CV + 0.6 × impression CV. Low CV = consistent output and predictable reach.

< 0.3100

0.3 – 0.575

0.5 – 0.850

0.8 – 1.230

> 1.210

4. Engagement Quality — HQ Interaction Ratio (15%)

HQ ratio = (replies + retweets + quotes + bookmarks) / (all interactions) × 100%. High ratio = real discussion, not just passive likes.

> 30%100

20% – 30%80

10% – 20%60

5% – 10%40

< 5%20

Domain Factor (10 Subcategories)

Claude AI identifies both the main category and subcategory to apply the appropriate CPM premium.

Domain × SubDomain → Multiplier (max 1.30x)

Crypto — DeFi / Layer1/L2 / Institutional1.30x

Crypto — NFT / Gaming / Memecoin1.20x

AI — LLM / Base Model / AI Infra1.30x

AI — Application / SaaS / Tools1.20x

Finance — Institutional / TradFi / Macro1.30x

Finance — Retail / Personal Finance1.10x

Business — Startup / SaaS / B2B1.20x

Tech — General / Developer1.20x

Entertainment — Lifestyle / Culture1.00x

Other1.00x

Other Modifiers

Credibility (AI-evaluated, max 1.10x)

Claude AI evaluates account authenticity: follower-to-engagement ratio, posting patterns, content originality, signs of manipulation.

85 – 1001.10x

70 – 841.00x

55 – 690.75x

40 – 540.50x

< 400.25x

Relevance (AI-evaluated, max 1.10x)

Claude AI judges each tweet's relevance to the account's domain. Measured as relevant tweets / total tweets.

85 – 1001.10x

70 – 841.00x

55 – 690.75x

40 – 540.55x

< 400.30x

Identity (AI-evaluated)

Builder accounts carry higher trust and conversion value.

Builder1.30x

KOL1.10x

Content Creator1.00x

Scarcity Factor (Ad Ratio, near 30 days)

Accounts that rarely accept sponsorships command a premium. Accounts that over-commercialize receive a discount.

< 5% sponsored tweets1.20x (very scarce)

5% – 15%1.10x

15% – 30%1.00x (baseline)

30% – 50%0.85x

> 50%0.70x (over-commercialized)