Project Score
A 20 to 100 algorithmic rating of every Canton Network project listed on CCTools. An AI reading of the disclosed fields and the profile's completeness set the base, on-chain activity adds up to 25 points, and serious operators on the Global Synchronizer skip the field-density cap.
The pipeline
- AI evaluation (60%). A model rates the project 0 to 100 overall, and on each of 5 dimensions, scoring only from the disclosed fields. The overall reading counts for 60% of the score. Brand recognition from training data is explicitly disallowed in the prompt.
- Completeness (40%). Weighted disclosure score totals 70 points across 35 fields, scaled to 0 to 100. It counts for 40% of the score.
- On-chain bonus. Up to +25 added to the final score. Validators, Featured Apps, and reward magnitude each add points, listed in the table below.
- Hard-cap defensive. Final score is capped by how many fields the project actually disclosed. Bypassed for serious on-chain operators with 5 or more validators or 10M CC or more in rewards.
The 5 dimensions
An AI model rates each dimension 0 to 100, scoring only from disclosed fields. For each dimension the model is told to keep its rating at or below a ceiling when the fields that support it are missing (the ceilings listed below). That is an instruction to the model, not a limit the code applies to its answer.
Displayed values. The value shown for each dimension on a project page is about AI_dim × 0.55 + completeness × 0.20, plus the on-chain points for that dimension, capped at 95. That 95 is the only hard limit on it. It has its own formula and is shown for reference: it does not add up or average to the score above.
Transparency
Public team identity, regulatory registration, audit trails, legal entity disclosure. Institutional-grade disclosure scores higher than anonymous projects.
The AI is instructed not to rate above 65 unless legal_name, entity_type, audit_url, or a named team is provided.
Activity
Live product with real users, app URL, on-chain validator or featured app presence.
The AI is instructed not to rate above 65 unless a working app or product URL, or on-chain activity (validators or Featured Apps), is shown.
Community
Real-world adoption: institutional clients, partnerships, team listed publicly. Social channels are a bonus signal, not the primary.
The AI is instructed not to rate above 70 unless team members, partnerships, or institutional clients are listed.
Documentation
Technical depth of docs site, whitepaper, API references, SDK. Evaluated for completeness and developer accessibility.
The AI is instructed not to rate above 60 unless docs, whitepaper_url, or a substantial description is provided.
Maturity
Track record: funding round, funding amount, lead investor, mainnet launch date, time on chain, accumulated rewards.
The AI is instructed not to rate above 70 unless funding_amount_usd, mainnet_launch_date, funding_round, lead_investor, or significant on-chain rewards are present.
On-chain bonus
On-chain presence is harder to fake than disclosed fields. Validators, Featured Apps, and accumulated rewards each add points, and the total is added to the final score, capped at +25.
All triggers stack, and the points from every row are summed. A project with 20 validators, 3 Featured Apps, and 10M CC rewards earns +15 Activity, +13 Community, +18 Maturity, 46 points in all. The bonus is capped at +25 on the final score; the displayed dimension values keep their own points, up to their cap of 95.
Hard-cap defensive
Independent of AI optimism, the final score is capped by how many fields the project actually disclosed. This catches halo bias where the model recognises a brand and inflates its overall reading past what the row supports.
Bypass. Projects running 5 or more active validators, or with 10M CC or more in rewards, skip the density cap entirely. Operators with real network commitment aren't penalised for withholding marketing details.
Completeness fields
Disclosure weight totals 70 points across 35 fields, grouped by importance. The list mirrors COMPLETENESS_FIELDS in the backend scoring prompt; keep in sync when the backend list changes.
Core substance
- Description (over 30 chars)5 pts
- Canton summary (≥ 30 chars)5 pts
High-value
- Documentation link4 pts
- Features (≥ 1)4 pts
- Team (≥ 1)4 pts
Substantive
- Funding amount (USD)3 pts
- App URL3 pts
Quality contributors
- Logo URL2 pts
- GitHub link2 pts
- Audit URL2 pts
- Twitter / X link2 pts
- Legal name2 pts
- Profile type2 pts
- Mainnet launch date2 pts
- Funding round2 pts
- Lead investor2 pts
- Whitepaper URL2 pts
- FAQ (≥ 1)2 pts
- Relationships (≥ 1)2 pts
- Connections prose2 pts
Network roles
- Super Validator role2 pts
- Validator role1 pt
Light signals
- Discord link1 pt
- Telegram link1 pt
- Open source = true1 pt
- Token address1 pt
- Entity type1 pt
- Token ticker1 pt
- Token type1 pt
- Privacy model1 pt
- Asset standards1 pt
- TPS target1 pt
- Finality target (ms)1 pt
- Supported countries1 pt
- Roles (≥ 1)1 pt
When the score is hidden
A score that doesn't reflect the project would do more harm than no score at all. Two cases hide the badge entirely across the ecosystem listing, the project page, the engagement strip, and the graph node detail panel:
- Status is "Announced". The project hasn't shipped a public surface to evaluate yet. Score returns automatically once status moves to Building or Live.
- Profile completeness under 50%. Less than half of the disclosed fields are filled. Score would be dominated by hard-cap defensive (≤ 35) and read as a punishment for new projects still being onboarded. Score returns automatically when an edit pushes completeness past 50.
Both conditions are non-permanent. Filling the profile or promoting status restores the badge on the next pageload, no rescoring required.