Choose Business Identifiers Before Joining Sales Data

Joining two sales datasets on company name can produce a tidy file with the wrong relationships. Similar names may represent separate businesses, while one business may appear under several names. Choose what an account represents before choosing a matching key. A legal entity, a trading brand, a store, and a corporate group are different units, even when they share a website.

Define the unit you want to match

Write a one-sentence definition such as “one record per separately registered operating company” or “one record per buying location.” Then inspect whether both source datasets use that unit. A group-level revenue estimate cannot be attached to every subsidiary as if it described each subsidiary individually. The join needs either a compatible unit or an explicitly labeled relationship.

Keep an internal account identifier that does not depend on a changeable name. External identifiers can support matching, but preserve their issuer and scope. GLEIF describes the LEI as a 20-character code identifying legal entities. That does not make it a universal identifier for every brand, branch, or prospect in your dataset.

Build a matching ladder

Start with exact verified identifiers when both sources contain them. Next consider a corroborated combination of company name, jurisdiction, and address. Use domains as supporting evidence rather than unquestioned proof of identity. A group domain may serve several entities, and a business may maintain multiple domains.

1. Confirm the entity definition in each source.

2. Check identifiers for blanks, duplicates, and inconsistent formatting.

3. Separate exact matches from candidates requiring review.

4. Inspect one-to-many and many-to-many relationships before merging fields.

5. Keep unmatched rows with a reason instead of forcing a match.

If you are comparing research options for identifier checks, ask about the US$15/month offer. Leads Automation is independent and not affiliated with LinkedIn, and this is not LinkedIn’s official price. Confirm the plan, eligibility, and terms before ordering.

Measure what the join changes

Record input row counts, distinct identifiers, matched rows, unmatched rows, and output rows. Investigate unexpected multiplication. If a table with one row per company joins to a table with several locations per company, extra rows may be correct, but the result no longer has one row per company. Label the new unit explicitly before reporting totals.

Do not overwrite the source files. Preserve a permitted, controlled input snapshot and a mapping file that records the chosen match and evidence. A reversible import pilot can test how those relationships behave in the destination before a larger update.

Fictional example: one brand, three entities

In this fictional example, Harbor Foods appears once in a marketing list and three times in an entity register. The marketing row describes the group brand, while the register lists three operating companies in different jurisdictions. A name-only join would attach the same group description and estimated scale to all three rows without explanation.

The researcher creates a group account with three linked entity records. Group-level observations remain on the group record; local addresses remain on the entities. One entity has insufficient evidence for a domain match and is marked for review. The output preserves uncertainty instead of hiding it inside a seemingly complete file.

Use exception review to improve the key

Look at patterns among unresolved matches. Repeated branch confusion means the entity definition may be wrong for the task. Repeated spelling differences suggest a normalization issue. Repeated identifier collisions suggest the supposed key is not unique or has been reused incorrectly. Repair the cause before lowering the match threshold.

The guide to safe company-name standardization can help prepare names for comparison without removing legal distinctions. Keep the original name alongside the normalized comparison value. A matching convenience should never replace the evidence about who the business actually is.

When reviewing the US$15/month offer, ask which identity checks, if any, are included. Do not assume a subscription can resolve ambiguous entities automatically or that it grants permission to combine every available dataset.

Approve a join only when someone can explain its unit, key, exceptions, and row-count changes. High match coverage is not a success if the matched records refer to different businesses.