Start with the record that drives the decision
Company data quality is not just whether a field is filled in. A complete address attached to the wrong subsidiary is still wrong. Define the business object first: a legal entity, a customer account, a supplier relationship or a corporate group. Each needs different matching rules.
For CRM and supplier master data, keep the internal record ID separate from the legal-entity identifier. Several accounts may legitimately refer to one company, and one group relationship may involve several legal entities. A cleanup process should preserve these distinctions.
1. Establish a reliable entity match
Use jurisdiction and registration number as strong matching inputs, then compare the legal name and supporting attributes. Names and domains are useful candidate signals, but they can change or refer to multiple entities. Keep uncertain matches in a review queue.
Measure confirmed matches separately from suggested matches. Otherwise a high match percentage can conceal incorrect joins that later spread through reporting, ownership analysis and customer communication.
2. Define fields before overwriting them
| Field pair | Why the distinction matters |
|---|---|
| Registered address / delivery address | A registry change should not overwrite a destination used for operations |
| Legal name / trading name | Both may be valid and useful in different workflows |
| Company country / sales territory | A jurisdiction is not the same as an internal commercial assignment |
| Reported employees / estimated employees | Observed filings and estimates have different meanings |
| Company revenue / group revenue | Consolidated scope can include several legal entities |
Assign an owner and an update rule to each field. Some values can be proposed from the external source; internal account assignments or negotiated commercial fields should remain under the business owner’s control.
3. Preserve provenance and dates
Attach a source reference and retrieval time to the external value. Preserve an effective or filing date when supplied. This allows a reviewer to answer whether a proposed update reflects a new record, an old filing collected recently or a mapping correction.
Do not make the newest retrieval automatically win every conflict. A newer response can be incomplete, or it can describe a different field definition. Retain both values and the reason for the selected value when a conflict needs review.
4. Resolve duplicates without merging away meaning
Separate duplicate legal-entity records from multiple valid business relationships. A customer and supplier record for the same company may need a shared entity reference while retaining separate workflow owners. Subsidiaries in the same group must not be merged into one company record.
Before a merge, identify dependent records, retained identifiers and the ability to reverse the change. For uncertain candidates, create a proposed linkage for review instead of deleting or merging immediately.
5. Run the company data quality checklist
| Dimension | Question to test | What to measure |
|---|---|---|
| Identity | Is the record linked to the correct legal entity? | Confirmed matches and unresolved candidates |
| Completeness | Are the fields required for this workflow present? | Required fields populated with usable values |
| Consistency | Do equivalent fields follow the same definitions? | Mapping conflicts and format exceptions |
| Freshness | When was the last successful source check? | Age of successful checks and unresolved failures |
| Traceability | Can the reviewer retrieve the source? | Updates with usable evidence references |
| Control | Was the change reviewed and applied correctly? | Approved, rejected and unresolved update proposals |
Choose a representative sample before automating updates. Include subsidiaries, renamed entities, conflicting addresses and records that cannot be matched. Accuracy on the straightforward records alone is not enough.
6. Turn changes into reviewed update proposals
- Compare a new successful company record with the saved baseline.
- Identify the external fields that changed and map them to eligible internal fields.
- Prepare a proposal containing old value, new value, source and reason.
- Let the responsible reviewer approve, reject or request further evidence.
- Record the outcome and confirm any delivery to the configured destination.
Keep delivery confirmation separate from approval. An approved update is not proof that the external system accepted it. Retried deliveries should preserve a stable reference so the receiving system can handle duplicates.
Where CompanyDelta fits
CompanyDelta supports the company-record portion of this workflow: company matching, saved snapshots, detected changes and investigation review. The company data management solution shows an example of a proposed field update.
External CRM or supplier-system updates require a configured integration and review policy. Begin with the fields that matter most to your workflow, prove the source mapping and make exceptions visible before expanding the scope.
Questions about company data quality
Is company data enrichment the same as data quality?
Enrichment adds or updates information. Data quality also requires that the information belongs to the right entity, uses the right definition, is sufficiently current and can be traced to a source.
Should we merge companies that share a website?
No. A shared domain can represent a group, several subsidiaries or a brand. Confirm legal-entity identifiers and business relationships before deciding whether records are duplicates.
How should we handle missing external fields?
Preserve known internal values unless your policy and evidence justify replacing them. Distinguish a missing field from a reported empty value and from a failed retrieval.