Playbook · 8 min read
How to build an enrichment waterfall — without corrupting your CRM
More data providers do not automatically create better data. A useful waterfall knows what it is trying to fill, which source wins, and when it should leave a field blank.
Enrichment is a decision system: resolve identity first, choose sources by field, preserve provenance, and never let a weak result overwrite stronger evidence.
What an enrichment waterfall is for
An enrichment waterfall sends a record through ordered sources until it finds an acceptable value. It is useful when no provider has complete coverage. It is dangerous when every result is treated as equally trustworthy.
The goal is not maximum field completion. The goal is fit-for-purpose data: enough confidence to route a lead, personalise responsibly, or report accurately.
Build the waterfall by field, not by vendor
| Field | What good looks like | Typical acceptance rule |
|---|---|---|
| Company domain | A canonical, active domain. | Matches the company website and is not a generic email host. |
| Employee range | A range suitable for segmentation. | Fresh enough for the sales motion; conflicts are flagged. |
| Industry | A controlled category the team can act on. | Maps to your internal taxonomy, not a provider’s raw label. |
| Person title | A current role at the matched account. | Has a source date and clears a title-normalisation rule. |
| Email status | A verified sendable address or an explicit no-send state. | Passes verification immediately before activation. |
The order of operations
- Resolve identity.
Normalise the company name and domain before enriching anything else. Do not enrich a contact against an ambiguous account.
- Check first-party data.
Your CRM, product data, form submissions, and customer conversations often know more than a third party. Treat them as the default authority where relevant.
- Query specialist sources.
Use sources that are strongest for the exact field you need, then apply a field-level acceptance rule.
- Verify before activation.
Validate sendability, recency, and constraints immediately before using data in an outbound or routing action.
- Log the result.
Save source, timestamp, confidence, and the reason a value was accepted or rejected.
Set overwrite rules before connecting anything
The most expensive enrichment failure is silent overwrite: an automated process replaces a confirmed value with a plausible but incorrect one.
| Existing value | New value | Recommended action |
|---|---|---|
| First-party, recent | Third-party, unverified | Keep the existing value. |
| Third-party, stale | Third-party, recent and verified | Update and retain both sources in history. |
| Blank | Accepted provider result | Write the value with provenance. |
| Conflicting high-confidence values | Another high-confidence result | Route to a review queue or choose a documented precedence rule. |
Track waterfall health
- Coverage: What share of eligible records receives an accepted value?
- Conflict rate: How often do trustworthy sources disagree?
- Freshness: How old is the evidence at the point of use?
- Cost per usable record: Are credits generating data the business can act on?
- Downstream outcome: Did the enriched records route better, convert better, or reduce manual work?
A practical starting configuration
Begin with one workflow: enrich accounts that meet your ICP, identify one or two role groups, verify email only for accounts that pass your relevance check, and send the results to review before activation. Once accuracy is stable, increase automation—not before.