People move, buildings get renumbered, and every form invites a typo. Most bad addresses in your database are not fraud. They are entropy, and they cost the same.
Not "is this a valid address". It is does this person live at this address, which is a different question and the one that matters for compliance.
Cleansing: formats normalised to the local standard, so records are comparable across countries.
Validation: does this address exist and is it deliverable.
Verification: is this person associated with it, from postal, official and private sources.
At entry: type-ahead capture, so the address is right when it is typed rather than corrected afterwards.
Where somebody has moved, relocation and deceased data resolves what the old address no longer answers.
Address verification in 50 countries.
A wrong address is a failed delivery, a returned card, a compliance record that does not hold, and a customer who is annoyed about something they may not have caused. Under AML rules it is also a verification you claimed to have performed against a fact that was not true.
Validate at entry, verify at onboarding, or both. Reject, correct silently, or ask, per market.
Address errors surface as four different problems, delivery, credit, AML, onboarding, each traced back to the same bad field.
What was submitted, what it resolved to, and how confident the match is.
Address is the least glamorous data product and one of the most load-bearing. It sits underneath identity verification, delivery, credit and AML at once, which means its errors surface as four different problems that nobody traces back to the address.