← Back to home

How scoring works

A structured methodology, not a single AI prompt.

Every project passes through a defined framework — maturity levels, scored dimensions, regulatory checks and human-verified facts — before it earns a STARTA Score.

1 · Technology Preparedness Level

Nine maturity stages, from concept to commercial product. Select a level to see what it means and how it's assessed.

TPL-1Idea

Concept defined — the problem and approach are articulated, but no code exists yet.

Assessed across 7 readiness criteria

Model maturity
Data maturity
Architecture maturity
Functionality maturity
Risk & compliance maturity
Testing maturity
Commercialization maturity

2 · Six scored dimensions

The overall 0–100 score breaks down across six areas.

Product Maturity

0

How developed and market-ready the product is, measured directly against its TPL maturity level (1–9).

Market Validation

0

Demand, traction, active users and revenue — the signals that show real market pull.

Technical Health

0

Architecture, stability and compliance — including EU AI Act readiness and system robustness.

Founder Profile

0

The team behind the project — experience, track record and verified identity.

Growth Signals

0

Growth rate and momentum, read from the project's own metrics over time.

Acquisition Fit

0

How attractive the project is to a buyer or investor — readiness for acquisition or investment.

3 · Regulatory & data layer

Checks a generic AI score never covers.

AI Act Readiness

Alignment with EU AI Act requirements.

  • Risk classification
  • Human oversight
  • Bias mitigation
  • Explainability & transparency
  • Safety & monitoring

Data Readiness

Quality and integrity of training data.

  • Quality & balance
  • Provenance & traceability
  • Anonymization & privacy
  • Bias & noise control

System Compliance

Robustness and reliability of the system.

  • Hallucination control
  • Model drift
  • Uncertainty handling
  • Architecture robustness

AI Act, Data Readiness and System Compliance apply to AI projects; non-AI projects are measured on general architecture, stability and security.

4 · Human-verified, not just generated

Facts are checked, not taken on trust.

Company registryFounder identityRevenue proofLive product

The result: a verifiable certificate

A PDF report with a unique ID and QR code — proof you can share with investors.