Address Quality Dashboard Shows 99.2% While Returned Mail Rises
A retail bank sends 1.8 million monthly statements. In Q2, 4.7% of statements were returned undeliverable, up from 1.1% a year earlier. The bank's data quality dashboard shows address completeness at 99.2%, so the data stewards say quality is fine. Operations suspects the dashboard: the completeness rule checks only that the address line 1 field is non-null, not that it is valid or deliverable. A sample of 500 returned statements found 38% had a postal code that did not match the city, 24% had an apartment number in the street field, and 11% had a country value of 'USA' spelled five ways. The bank is launching a new customer onboarding API in six weeks and plans to reuse the same address validation logic. The vendor address-cleansing tool costs $180,000 annually; a one-time manual cleanup is estimated at 22,000 hours. The CDO must decide what to fix first and whether to delay the API.
1. Why is the 99.2% completeness metric misleading, and what measurement approach should replace it?
Read the worked answer
Completeness answers only whether a field is populated. It does not answer whether the address is valid, conforms to a standard, is consistent across systems, or is deliverable. The business outcome - 4.7% undeliverable mail - is the fitness-for-purpose measure, and it contradicts the dashboard.
Replace the single metric with a small scorecard tied to the business process:
- Deliverability rate: percentage of mailed statements not returned, measured monthly and segmented by acquisition channel, legacy migration batch, and country format.
- Validity/conformity: postal code-to-city match, state code conformity, street suffix standardization, and apartment-number placement.
- Consistency: same customer address across CRM, core banking, and statement system.
- Accuracy: sampled verification against a postal authority or vendor reference file.
Operational definitions matter: for example, 'valid address' should mean it passes the postal validation service and is accepted for mailing, not merely non-null. Sample sizes should support confidence for each segment. The dashboard should show the outcome metric alongside root-cause dimensions so stewards can see which upstream process is producing defects.
2. Recommend a remediation sequence and state whether the onboarding API should be delayed.
Read the worked answer
Sequence by risk and by whether the fix prevents future defects or cleans existing ones:
- Contain the mailing problem now: run the highest-risk segments - returned mail, new accounts from the past 12 months, and non-standard country values - through the vendor cleansing tool. Do not wait for a perfect enterprise rule.
- Fix the point of entry: the API should not reuse a rule that only checks non-null. Embed postal validation, standardization, and a reject-or-queue path for invalid addresses. This is the root-cause fix.
- Remediate the legacy base in waves: prioritize customers with active statements and high-value relationships; 22,000 manual hours is too large for a single cleanup, so use vendor matching plus targeted human review.
- Decide the API date conditionally: delaying the API by two to four weeks to embed validation is cheaper than launching a new defect source at scale. If the business refuses a delay, launch behind a feature flag that blocks unvalidated addresses from statement mailing and queues them for review.
The $180,000 annual vendor cost is justified if it prevents even a fraction of 1.8 million monthly statement failures. The manual cleanup estimate should be revisited after profiling, because not all records need human touch.
3. Who should own address quality, and what service levels would you set?
Read the worked answer
Accountability should sit with the business process owner - Customer Operations - because address quality affects statement delivery and customer contact. The data steward owns rules and definitions; IT owns the validation service and integration; the CDO's office owns measurement and escalation.
Service levels should be process-based, not field-based:
- At onboarding: at least 98% of new customer addresses pass postal validation before the account is activated; invalid addresses enter a review queue within one business day.
- At statement mailing: at least 99.5% of statement addresses are deliverable, measured by return rate; a return rate above 2% triggers root-cause review.
- Defect correction: address defects in active customer records corrected within 10 business days for high-value segments and 30 days for the rest.
- Reporting: monthly DQ scorecard to the data governance council with segmented return rates and top defect sources.
This assigns ownership where the outcome is felt, while keeping data management accountable for rules and measurement. It also prevents the stewards from defending a completeness metric that does not reflect operational performance.