Official AIMI Score v1.0 Engine

The normalized scoring engine behind every AIMI Score.

Step 4 of the AIMI Ledger backend roadmap. The Full Normalized AIMI Scoring Engine converts raw platform signals into a transparent 0–1000 score using cohort normalization, weighted components, and anti-gaming multipliers.

Beta / Prototype. The scoring engine architecture is documented here for transparency. Live external platform data connections are in progress. Scores shown on sample profiles use representative demo data until the FastAPI engine is deployed.

Step 4 · In Progress

Full Normalized Scoring Engine·

Admin Dashboard: Complete

Scoring Engine: In Progress

Prompt Novelty (pgvector): Step 5

Provenance Layer: Step 6

Official Signal Weights

Five weighted signals. One score.

Every AIMI Score is composed of five normalized signal components. These weights are the official v1.0 configuration, versioned as 2026.1.

30%

Streaming Velocity

Normalized streaming volume and week-over-week growth velocity across verified platform APIs.

25%

Social Engagement

Likes, shares, comments, and saves aggregated across social platforms, normalized by cohort.

20%

Prompt Innovation

Prompt uniqueness, novelty score, and template-spread velocity from the prompt registry.

15%

Media Footprint

Editorial mentions, press coverage, and playlist placements from verified editorial sources.

10%

Community Signal

Community votes, upvotes, and qualitative engagement signals from the AIMI community.

Normalized Scoring

Raw counts are never used directly.

Cohort Percentile

Each signal is normalized into a percentile within its chart-week cohort. A track with 10,000 streams in a slow week may rank higher than 50,000 streams in a blockbuster week.

H-Level Tier Grouping

Normalization happens within H-Level cohorts. An H3 track is compared against other H3 tracks, not against H0 human-created or H5 fully-synthetic works. This ensures fair comparison.

Chart-Week Freshness

Each normalization cycle is tied to a chart week. Older snapshots decay, so the score reflects current market influence, not lifetime accumulation.

Score Formula

The AIMI Score equation.

AIMI Score v1.0 · Methodology 2026.1

AIMI Score=1000×weighted normalized signals×freshness×platform diversity×data confidence

Weighted Signals: Σ (component × weight)

Freshness: 0.5–1.0 decay

Diversity: 0.7–1.0 penalty

Confidence: 0.3–1.0 tier

Metrics Snapshot Architecture

Raw signal capture schema.

Each chart week, the engine captures a Metrics Snapshot per track — the raw normalized inputs before weighting and multiplier application.

MetricsSnapshot

10 fields

streaming_velocity

float

Normalized streaming volume for the chart week

social_engagement

float

Normalized engagement score (likes, shares, comments)

prompt_innovation

float

Prompt novelty score from the prompt registry

media_footprint

float

Editorial mentions and playlist placements count

community_signal

float

Community votes and qualitative engagement score

platform_distribution

json

Distribution of metrics across platforms (diversity)

source_confidence

float

Weighted confidence from source tier classification

captured_at

timestamp

When the metrics snapshot was captured

chart_week

string

ISO chart week identifier (e.g. 2026-W26)

h_level_tier

enum

H0–H5 cohort tier for normalization grouping

Score Record Architecture

Final score calculation schema.

The Score Record stores the weighted component scores, anti-gaming multipliers, and the final AIMI Score (0–1000) with its methodology version.

ScoreRecord

11 fields

streaming_velocity_component

float

Weighted component score (30% weight)

social_engagement_component

float

Weighted component score (25% weight)

prompt_innovation_component

float

Weighted component score (20% weight)

media_footprint_component

float

Weighted component score (15% weight)

community_signal_component

float

Weighted component score (10% weight)

freshness_multiplier

float

Decay multiplier — recent data scores higher

diversity_multiplier

float

Penalizes single-platform dependency

confidence_multiplier

float

Scales by source data confidence tier

final_score

integer

Final AIMI Score (0–1000) after all multipliers

methodology_version

string

Scoring methodology version (e.g. 2026.1)

calculated_at

timestamp

When the score was calculated

Anti-Gaming Methodology

Built-in manipulation resistance.

Five structural safeguards ensure the AIMI Score reflects genuine market influence, not artificial inflation.

Freshness Decay

Older metrics lose weight over time, keeping charts responsive to current performance rather than historical accumulation.

Platform Diversity

Tracks relying on a single platform receive a reduced diversity multiplier, discouraging single-platform manipulation.

Source Confidence

Unverified and manual-claim data sources carry lower confidence multipliers, limiting their influence on final scores.

Cohort Normalization

Raw counts are never used directly. Each signal is normalized into a percentile within its chart-week and H-Level cohort.

Admin Review

Verification status is reviewed by human admins and kept separate from score rank, protecting trust quality.

Operational Loop

From submission to chart eligibility.

Step 1Submit Track

Step 2Assign AIMI ID

Step 3Admin Review

Step 4Capture Metrics

Step 5Normalize Signals

Step 6Calculate AIMI Score

Step 7Generate Trust Profile

Step 8Chart Eligibility

Explore the full methodology.

Review the complete AIMI Score v1.0 framework, the backend engine roadmap, or a sample trust profile showing the score breakdown in action.

Full Methodology Backend Engine Roadmap Sample Trust Profile