Skip to main content
Workflow demos · Finance

Validate credit scoring models
and calculate flood risk to collateral

Recomputes a credit scoring model's discrimination (AUC, KS), stability (PSI), rank ordering and calibration from its scores and the defaults that actually followed, and produces a validation metrics table, charts and a validation report draft. Overlays the locations of collateral properties on a flood hazard map (SHP) to calculate exposure by hazard grade, and also produces a map and a climate risk report draft.

Try it with your own dataLearn about harnesses
Real product screen
Collateral over the hazard zones
01

Task overview

Use it for periodic validation of a credit scoring model from each borrower's score and grade and whether the borrower defaulted after the observation period, or to calculate the flood risk exposure of collateral from a collateral list (address, appraised value and loan balance) and a flood hazard map. The results are drafts; the risk management department makes the final decision.

StepsMap columns → recompute metrics → overlay map → report draft
RuntimeOffice documents image · up to 30 steps
Completion checks2 check scripts · Excel · charts · report draft
02

Input

Sample data

Attachments

scores-defaults-2025.csv + 3 files

Request

Validate last year's credit scoring model, and also summarize the flood risk exposure of our collateral properties.

Company data to connect

  • Borrower scores, defaults, collateral and loans

    Used for the full analysis: each borrower's score, grade and predicted default rate, whether the borrower defaulted after the observation period, and the collateral list (address, coordinates, appraised value and loan balance). When not connected, only the input files are analyzed.

  • Model risk and climate risk standards

    The bank's model validation thresholds (AUC and KS floors, PSI alert levels and the significance level for the binomial test) and its flood hazard grades. When connected, the bank's standards are applied instead of the tool defaults and quoted in the answer.

03

Processing steps

  1. Connect the risk database

    Copy the template and connect the risk database.

  2. Upload the development sample

    Upload the development sample file in chat, and discrimination and stability are recomputed on all borrowers in the risk database.

  3. Review the validation results

    You get a validation metrics table, charts and a validation report draft.

  4. Analyze flood risk

    Then upload a flood hazard map and the collateral list to get the loan balance by hazard grade and an exposure map.

04

Output

Result files

model-validation.xlsxroc.pngflood-exposure.xlsxflood-map.png

Completion check items

  • Metrics and default rates by grade recomputed
  • Hazard grade re-evaluated for each collateral property
  • Metrics table, charts and report draft
05

Sample run

Start from a harness templateStart by copying a domain's instructions, skills and completion checks, then connect your company's data to the resources the template recommends.

Risk managementSample run
scores-defaults-2025.csv + 3 files

Validate last year's credit scoring model, and summarise how much of our collateral sits in flood zones.

Completion checks
Result files

A sample screen. Real steps and times depend on the model and the files.

06

Sample data

All file names and figures are sample data.

Sample data

Validation metrics and exposure

MetricValueStandard
AUC (all)0.7810.70 or above
KS0.430.20 or above
PSI vs development0.14Watch 0.10
Grade order1 inversion (7–8)No inversion
Collateral in flood hazard grades 1–237 · KRW 41.2bn—
48,210 borrowersV1 · V2 · H1

See it with
your own documents

In a consultation, we connect a few of your documents so you can check the answer quality for yourself. Send us your questions and our team will get back to you.

Try it with your own dataAll workflow demos
Product videosReal product screens · Example company ACME Logistics · 16 videos