Professional work · Data architecture · identity · governance
Many records. One customer picture.
A shared identity foundation for sales, marketing, fraud, and applied AI.
Walmart Global Tech · customer identity
Same name without case, punctuation or legal suffix, in the same city.
8 records resolve to 4 organizations. Resolved.
Eight records, four organizations
Ignore case, punctuation and legal suffixes, and require the same city. The three Northwind Traders records become one organization, and Northwind Dental stays its own business.
The golden record keeps every useful address: the street office and the mailing box.
Too eager: two businesses become one
Matching on the first word alone pulls a dental clinic in Tacoma into Northwind Traders. Sales would call the wrong company, and fraud checks would mix two histories.
An over-merge is the expensive mistake. It is hard to see and harder to undo.
Too careful: one business becomes two
Requiring the same street address splits off the record that only has a mailing box. Its signals now belong to a second Northwind Traders that nobody owns.
An under-merge is quieter. It shows up later as a missed signal or a duplicate call.
Identity includes forgetting
A deletion request removes one source record. The organization survives, but the golden record is rebuilt from what remains, so the mailing box it came with disappears too.
Matching, retention and deletion are one system. A golden record has to stay true as its sources change.
Your rule
Each rule is a trade. Looser rules join more records and risk merging different businesses. Stricter rules protect against that and risk splitting one business in two.
The right rule depends on which mistake costs the teams using it more.
Illustrative companies and addresses. The rules are simplified to show the two ways matching goes wrong.
- distinct organizations
- 1.8M
- sales, marketing & fraud stakeholders
- 150+
The same company, written five ways
Northwind Traders Inc. in one system. NORTHWIND TRADERS in another. Northwind Trading Co. at a new address after a move. Every team had its own version, and they did not agree on how many customers the business had.
One golden record
I built the customer identity foundation that brings those pieces together. Tens of millions of source records resolve into 1.8 million distinct organizations, and more than 150 people across sales, marketing, and fraud work from that shared picture.
It is also the base the go-to-market agents stand on.
The judgment: where to draw the line
Matching is a dial, not a switch. Set it loose and two different businesses with similar names become one customer. That mistake is worse than a duplicate, because nobody notices it. Set it strict and the same company stays split by an abbreviation or an old address.
Try the strictness control above. The right setting keeps both mistakes rare, and the cost of each one decides which way to lean.
Identity is a lifecycle
A record keeps changing after its first match. Companies move, merge, and ask to be forgotten. I treated retention and deletion as part of the same system as matching: when a source record has to go, the shared picture has to change with it.
Making it useful
Industry classification, lead segmentation, and routing sit on top of the identity. They answer everyday questions: which team should follow up, what an account needs, and where an opportunity belongs.
Two teams looking at the same customer should see the same customer.
Let’s talk ↗