Free tools / Bureau scrub
Bureau data scrub prompt for credit pulls
Checks a stored bureau extract for problems before you derive variables or build a model. It is free, needs no sign-up, and works from table and column names alone.
What you enter
You type names only: table and column names, never customer data.
- Bureau table
- Customer id column
- Application id column
- Pull date column
- Application date column
- Score column
- Valid score min
- Valid score max
- Stale after (days)
- Other columns to check
How to use it
- Use the extract you actually feed into underwriting or modelling.
- Run the summary query and sort by failure rate.
- Decide for each failure: fix at source, exclude, or impute. Write the decision down.
Checks to run on the answer
AI can write SQL that looks right and is wrong. Run these checks before you trust the result.
- Total rows in the checks equal total rows in the table.
- Special codes were counted apart from real scores.
- Duplicates were examined before anything was deleted.
- Each exclusion rule is written down with its count.
- Rerun after the fix and compare the failure rates.
Open the tool
You get the ready prompt, the checks above, what healthy and warning results look like, and how to present the result to an approver.
Open the free toolsRelated guides
Gross NPA vs bureau 90+: why the numbers never match
Gross NPA ratio and a credit-bureau 90+ rate look like the same thing. They are not. Learn the six reasons they differ before you compare them.
Using AIHow to check AI-written credit-risk SQL before you trust it
AI tools write SQL that runs but may be wrong. Ten checks to run on any credit-risk query: row counts, duplicates, DPD edges, bucket totals and a hand trace.