Our Assessment Approach

The numerical score is not shown in the panel or on the report cover; it appears only in the methodology section of the report. In a model based on expert judgement, small score differences are not meaningful; the decision is made on the class and the conditions.

1. 1

   Answers and documents

   Missing or unscorable mandatory answers are imputed as zero and remain missing in information sufficiency. Unanswered optional questions are excluded. Not applicable requires a reason and expert acceptance.

2. 2

   Dimension level

   Each dimension's level is calculated using pilot question weights: Strong, Adequate, Limited or Weak. If a critical dimension cannot be assessed, no class is given.

3. 3

   Weighted result

   Dimensions are combined using weights set for each service, producing the proposed class. The weight of a non-critical dimension that cannot be assessed is redistributed to the others.

4. 4

   Cap rules

   Serious findings are not averaged away: a triggered rule caps the best class that can be reached.

5. 5

   Expert assessment

   The GRC expert may adjust dimension levels and the class within limits, with written reasons. Positive intervention requires valid evidence references. Net or gross positive uplift above 5 points, or total uplift above one class, requires approval. Total uplift cannot exceed two classes. Evidence/scope and insolvency-process caps cannot be removed.

6. 6

   Approval and delivery

   After a separate approver's check, the report is locked as a version and published with a verification code. Any later change is made in a new version.

The role of AI: uploaded documents go through an AI pre-review, and report texts may be drafted with AI support. The score, dimension levels and caps are not calculated by AI; they follow the rule-based steps above. Critical documents cannot be automatically accepted and require expert review. Class A requires documentary review and high information sufficiency. Source-checked scope relies on recorded human checks of specified documents, not automatic official-source lookups. Unsubmitted or expired collateral earns no points; foreign-currency collateral requires a sourced FX quote and eligibility review.

Active moratorium and insolvency processes may impose a D cap; this is not automatic commercial rejection. If two separate serious findings are triggered together, the expert must consider a D class and justify the decision. A concern about a document is not reported as an accusation.

Model thresholds are initial values and are calibrated through a pilot study on past files. Every model change is a new version; reports already delivered stay on their own version.