Data quality strategy: lending data that credit decisions can trust
Aldermoor Bank · Chief Risk Officer and executive risk committee · 90 days
Approve a 90-day SME borrower data pilot, name the accountable data owner, and agree that thresholds are set with credit and risk users.
Summary
Run a 90-day pilot on one data product, the SME borrower profile, tying every quality check to a credit or reporting decision, a named owner and a correction route back to origination. The goal: reliable sector-exposure reporting without a new platform, and evidence on whether to extend the approach.
The question
Can credit officers and risk analysts rely on an SME borrower’s identity and industry code at the moment they decide or report?
Context
- Internal audit found inconsistent industry codes on SME borrower records.
- Relationship managers re-key details across origination, CRM and core banking; nobody owns the shared fields.
- Risk and finance reconcile the sector-exposure report by hand at each month-end.
Objectives
- Define SME data quality by the credit and reporting decisions it supports.
- Route each material issue to a named owner and fix it where the data is captured.
- Show with evidence whether reconciliation effort and repeat issues fall.
Scope
- SME borrower profile: identity, registered number, industry code, group links
- Origination capture, CRM and the sector-exposure reporting feed
- Credit, relationship management, SME operations and risk reporting
Out of scope: retail and mortgage data; new platforms or off-cycle core banking changes
Principles applied from the source
- 01Define quality for a specific use and decision
Judge industry codes by whether exposure limits and credit decisions can rely on them, not by a completeness score.
- 02Follow problems back to where data is captured
Trace coding errors to origination and give relationship managers weekly, non-punitive feedback.
- 03Give every issue an owner; judge fixes by evidence
Log each issue with an owner and correction route, then check whether rework and repeat issues fall.
Plan
- Days 1–30Define and baseline
- Map decisions back to source fields
- Write consequence statements with credit and risk users
- Baseline issues by count and exposure value
- Days 31–60Check and route
- Build four to six named checks; “valid format” is not “verified”
- Open an issue log with owners and a route to origination
- Agree a threshold for each consequence
- Days 61–90Test and decide
- Run the loop through a month-end close
- Review lasting fixes and repeat issues with users
- Recommend whether to extend to other lending data
Proposed measures — four lines beside each
| Measure and proposed target | Decision it supports | If it fails | Action triggered | Owner |
|---|---|---|---|---|
| SME exposure with a verified industry code ≥ 98% by valueBaseline in weeks 1–3; report count and value by segment. | SME credit approvals and sector limits | Exposure is booked to the wrong sector | Steward reviews the affected segment | SME operations lead |
| Working days from detection to source correction ≤ 10 daysMeasured from the issue log from phase 2. | Whether fixes keep pace with new issues | Backlog hides at month-end | Data owner reviews capacity | Head of SME Lending |
| Month-end reconciliation hours Baseline firstRecord the next close before setting a target. | Whether the pilot reduces manual rework | Effort stays hidden and unfunded | Set a target after the next close | Risk reporting lead |
| Corrected issues recurring within 60 days < 5%Shows whether fixes happen at source. | Whether fixes are made at source | The same errors return next month | Trace recurrence back to capture | Data quality lead |
Targets are proposals for owners to agree after the baseline.
Ownership
| Chief Risk Officer | Sponsor. Owns the decisions; signs off thresholds. |
|---|---|
| Head of SME Lending | Data owner. Definitions, capture standards, fixes at source. |
| SME operations lead | Steward. Weekly triage, corrections, feedback to relationship managers. |
| Data quality lead | Builds checks, publishes counts and rates, keeps the issue log. |
| Risk reporting lead | Data user. Confirms fitness for use at each month-end. |
Risks and responses
- Fixes land in the warehouse, so issues return. Every fix records a source action and an owner.
- Borrowed thresholds misjudge credit consequences. Set thresholds with credit and risk users.
- Core banking changes wait for the quarterly release. Start with origination guidance; queue system changes early.
Assumptions and open questions
- One industry classification standard is authoritative for SME lending.
- SME operations can give the pilot about one day a week.
- Which risk-data and reporting expectations apply? For qualified review.
- Who settles classification disputes between lending and risk?
Next steps
- Week 1: confirm the data owner and steward
- Weeks 1–2: run two consequence workshops
- Day 30: publish the first baseline, by count and value