Onboarding a CRM File With a 30% Match Rate
A subscription business onboards a monthly CRM extract of 2 million customer records via an inbound S3 data source, using hashed email as the only identifier. The first run matches only 30% of records to known Audience Manager users. The CRM team insists the file is clean because it validates internally. The marketing team wants to expand reach by uploading a purchased third-party email list to boost match rates. Compliance has flagged the purchased list. The account has a contractual obligation to keep PII handling auditable.
1. Diagnose plausible causes of the low match rate before changing anything.
Read the worked answer
A 30% match rate is low but not automatically a data-quality failure - it depends on the identifier and the population. Work through candidates:
- Hashing mismatch. The most common cause. Confirm the exact normalization applied before hashing: lowercasing, trimming whitespace, handling of plus-addressing, and the specific hash algorithm. If the CRM hashes with different normalization than Audience Manager expects, matches fail silently.
- Identifier coverage. Email-only matching depends on the destination-side identity graph having that email. If many customers signed up via app or social login, their email may never be tied to an Audience Manager ID.
- Audience overlap reality. 2M CRM records may simply include many people who have never been seen by the platform. A 30% match on a cold list is plausible.
- File formatting. Check delimiter, encoding, header handling, and whether the file was processed fully vs. partially rejected.
Sequence: validate normalization and hashing first, then check processing logs for rejects, then reason about population overlap. Do not assume the CRM is wrong or right - test each layer.
2. How should you respond to the request to upload a purchased third-party list?
Read the worked answer
Decline the purchased list as a match-rate fix, and explain the reasoning rather than just citing policy:
- Consent and provenance. Purchased lists typically lack the consent basis the original customers gave. Using them to expand reach creates a compliance exposure that a match-rate improvement does not justify.
- It treats a symptom, not the cause. If the real issue is hashing normalization, a purchased list inflates volume while leaving the underlying pipeline broken - and pollutes the audience with low-intent records.
- Auditability. The contractual obligation to keep PII handling auditable means every onboarded source needs a documented lawful basis and data source configuration that traces back to it. A purchased list is hard to document defensibly.
Offer alternatives: fix normalization, add a second consented identifier (e.g., hashed phone or a customer ID) to improve match, and onboard additional first-party sources. If the business still wants third-party data, route it through compliance for an explicit decision rather than making it an analytics workaround.