What is entity resolution?

How does entity resolution work?

Entity resolution standardizes records, groups the ones that could match, compares them, and merges each group into one record with a stable ID. Whatever the tool, from a spreadsheet formula to a dedicated engine, the four steps are the same.

1. StandardizeClean every recordLowercase, strip https://, www. and paths, drop Inc. or Ltd.
2. BlockGroup possible matchesBy domain or name prefix, so not every pair is compared
3. MatchCompare the candidatesAn exact key, or a closeness score above a threshold
4. MergeOne record per entityA stable ID, linked to every source record

The Stripe example above runs steps 1 and 3 through an API. Five spellings went in, one company page came out, and the record behind it carries a page ID that every later lookup can reuse. Resolution comes before data enrichment: you can only add facts to a record once you know which company it is.

Entity resolution vs identity resolution: what is the difference?

Entity resolution and identity resolution are the same idea, used on different things by different teams. Entity resolution matches any real-world entity, companies included. Identity resolution matches people, with the keys marketing holds.

	Entity resolution	Identity resolution
What it resolves	Any real-world entity: companies, people, products, places	People, usually customers or website visitors
Typical keys	Domains, company pages, legal names, profile URLs	Emails, phone numbers, logins, device IDs, cookies
Where you meet the term	Data engineering, master data management, B2B data	Marketing, customer data platforms, advertising
Result	One record per entity, linked to its sources	One profile per person across devices and channels

Identity resolution is entity resolution applied to people, with marketing’s keys. In B2B data the two meet: a person is resolved to a profile, and the profile is linked to a company that has been resolved in turn. Record linkage is the statisticians’ name for the same problem, from the long practice of matching census and health records.

Why is a domain not enough to resolve a company?

A domain is not enough to resolve a company. Even the strongest key a company has fails in three ways, and the evidence above shows all three.

• Spelling. stripe.com, www.stripe.com and https://stripe.com/en-us are one key once standardized. Left raw, they become three accounts.
• Shared claims. Besides Stripe’s own page, each lookup listed other company pages that point to stripe.com. A resolver that took the first page it found could merge them with Stripe.
• Similar names. Mercury, listed under Financial Services in San Francisco, and Mercury Insurance, listed under Insurance in Brea, share a word and a state. A match on name and state would merge two companies; their domains keep them apart.

Brands add a fourth trap: one company can run several domains for its products or its countries, all tied to one company page. That is why resolution starts from the domain and ends on the company page.

Which is which?

Five pairs of CRM records. Do they describe the same company?

1. acme.com and https://www.acme.com/about

2. Acme Inc. and ACME Corporation, both on acme.com

3. Globex at globex.com and Globex Logistics at globexlogistics.com, each with its own company page

4. Northwind, before and after it renamed itself Northwind Health, with the same domain and the same company page

5. Initech in Austin and Initech in Boston, with different domains and different industries

Show the answers

1. acme.com and https://www.acme.com/aboutSame companyStrip the scheme, the www and the page path: both reduce to the domain acme.com.
2. Acme Inc. and ACME Corporation, both on acme.comSame companyCapitals and legal suffixes change; the shared domain does not.
3. Globex at globex.com and Globex Logistics at globexlogistics.com, each with its own company pageDifferent companyA shared word is not a shared identity: two domains and two company pages make two entities.
4. Northwind, before and after it renamed itself Northwind Health, with the same domain and the same company pageSame companyA rename changes the label, not the entity: the keys stay the same.
5. Initech in Austin and Initech in Boston, with different domains and different industriesDifferent companySame name, different keys and different facts: merging them would mix two customers.

Entity resolution, deduplication and data matching: what is the difference?

Data matching is the comparison, deduplication is the cleanup inside one table, and entity resolution is the whole job across your CRM, your product database and every list you import. It is also the first step of data hygiene: duplicates cannot be merged until you know they describe the same company.

	What it does	Scope
Data matching	Compares two records and decides whether they describe the same thing	One pair at a time
Deduplication	Finds and merges repeated records	Inside one dataset
Entity resolution	Matches, merges and links records into one entity per real-world thing	Across every source

How do you resolve company records with an API?

To resolve company records with an API, map every record to a key the source recognizes before doing anything else.

Domain to LinkedIn turns a domain, or a full web address, into the company page that BlitzAPI’s other endpoints take as input, and lists the other pages that claim the same domain in a separate field. From the page, Company Enrichment returns one record with a stable page ID that later calls can reuse. For people, a work email plays the same role: a reverse email lookup resolves it to a profile and a current employer. The data resolution use case shows the full chain on a messy import.