What Is Entity Resolution?

Entity resolution is the ability of teams to determine that records located in different systems represent the same "real-world" entity, and to connect or bring those records into a single and correct representation. It cleans up enterprise entities such as customers, suppliers and companies, products, providers, etc., that are scattered, using entity matching, record matching, record linkage, similarity scoring, probabilistic matching and data deduplication.

Most data from enterprise sources is not uniformly identified and available with a single reliable ID that is used everywhere.

A customer could be listed in the CRM with one name, with another name in the ERP and with a third name in billing. Different addresses may be carried by a supplier depending on the region. The same company may have several short trading names in several systems, as well as be a parent company and a subsidiary in the same system.

Relying on loading everything into a lakehouse doesn't automatically mean that it shows which records are of a given entity. That is where entity resolution comes into play.

It provides enterprise data teams with an actionable approach to transition from "MESSY" data from source records to "ORGANIZED" master data. The process determines which records are linked to an entity, the extent to which these records are the same entity and what to do when the records are linked.

What Is an Entity Resolution?

Entity resolution is the activity of identifying, matching, and reconciling records that are all actually associated to a single entity in the real world (such as a person, company, supplier, product, provider, account, or another entity).

Those records do not match exactly, and this is where the challenge is. Look at these three company records:

• Acme Corporation
• ACME Corp.
• Acme Corporation USA

Identical names do not necessarily mean the same company. Even addresses, domains, tax IDs, phone numbers, parent company information and more still need to be checked.

A working entity resolution process will then evaluate a number of signals to make a decision on whether to merge records.

This is why entity resolution is related to entity matching, record matching, record linkage, identity resolution, and data deduplication, but is more than just a data deduplication of duplicate rows.

The goal is to create a solid image of the real-life person or thing represented by the records.

Why Fragmented Enterprise Data Is Difficult to Match

Systems that support businesses are created to help with various business processes. They were never designed to keep one clean, shared entity model.

CRM focuses on account details. ERP focuses on legal and financial data. Procurement systems handle suppliers in their own way. Third-party data often adds its own identifiers and extra information. As data moves between these systems, the inconsistencies keep growing.

Names and Attributes Change Across Systems

Even small differences can stop exact matching from working. Abbreviation, punctuation, spelling errors, legal suffixes, and regional naming conventions are all picked up by the company names. People move from place to place and change their phone numbers. Product attributes get written in different formats across catalogs.

Record matching has to handle these variations without wrongly joining records that do not belong together.

Important Identifiers May Be Missing

When available, strong identifiers like tax IDs, account numbers, or registration numbers can simplify the matching process. There are, however, no guarantees that they will always be available.

Some files only hold a name and address. Others carry incomplete contact details or outdated information. Entity resolution therefore has to work even when the best identifiers are missing or incomplete.

Different Entities Can Look Similar

Matching too aggressively creates the opposite problem: false positives. Two different organizations can share similar names. Family members can share an address. Suppliers can operate from the same office. Product records can carry nearly identical descriptions. The process must separate genuine duplicates from records that only look alike on the surface.

Enterprise Scale Changes the Problem

Matching a few thousand records is one thing. Resolving millions of records across multiple systems is a completely different challenge. Checking every record against every other record quickly becomes too expensive in computing power. Enterprise approaches need smart candidate generation, scoring methods, clear matching rules, and practical review steps.

How Entity Resolution Works

Most modern entity resolution workflows consist of the following steps: preparation, candidate identification, matching, confidence scoring, governance and consolidation. The details change by domain, but the overall flow stays consistent.

1. Standardize and Profile the Data

Records cannot be compared properly until the platform first understands the source data. Names get normalized. Addresses get standardized. Phone numbers, domains, dates, and identifiers get put into consistent formats.

Profiling also reveals missing values, weak attributes, unusual patterns, and existing duplicate clusters. Standardization alone does not resolve entities. It simply makes the later comparisons more useful.

2. Generate Candidate Matches

At enterprise volume, comparing every possible pair is not practical. Candidate-generation methods narrow the field to records that have at least some chance of referring to the same entity.

Those candidates might share:

• Similar company names
• Matching email domains
• Comparable addresses
• Common phone numbers
• Shared external identifiers
• Related geographic information

The matching engine then runs the heavier analysis only on this smaller set.

3. Compare Records Across Multiple Attributes

The system measures how closely the candidate records line up. Not every attribute carries the same weight.

A shared business identifier often counts for more than a similar company name. A matching email address can be stronger evidence than matching cities. Each field is significant in relation to the type of entity and business context. This is when the various matching methods can be utilised.

Deterministic, Probabilistic and AI-Assisted Entity Matching

There is no one single technique that is absolutely perfect for every enterprise data set. Most modern systems therefore combine several methods.

Deterministic Matching

Deterministic matching follows clear, fixed rules.

One example:

Company name + postal code + phone number must match.

These rules are easy to explain and predictable. They are effective where there are identifiers and fields that are consistent and reliable across the data.

These are, however, inflexible. If there is a typo, missing value, or varying formatting in a real match, then the match cannot be played.

Similarity and Fuzzy Matching

Similarity methods examine the "similarity" of values rather than requiring values to be equal.

They can flag records such as:

Global Technologies Incorporated and Global Tech Inc.

as possible matches.

Fuzzy matching is particularly useful for names, addresses, descriptions etc., fields that are naturally variable.

Probabilistic Matching

Probabilistic matching combines multiple pieces of evidence and attempts to determine the probability of matching two records.

It is not based on a simple "yes or no" rule. It can add strong and weak signals together. This flexibility can be beneficial for enterprise data that is already complex and for which exact rules are not sufficient.

AI-Assisted and Semantic Matching

AI can also detect semantic similarity that can't be detected by string comparison. On Databricks, the enterprise matching workflows are combined with AI techniques in LakeFusion. This MDM architecture combines similarity and AI-assisted inferences based on vectors for understanding disjointed enterprise entities, yet still enables human oversight if they are unsure.

The idea is not to train a big language model on each and every pair. Deterministic Rules, Probabilistic Scoring, Semantic Matching and Controlled Human Escalation based on the type of decision and the confidence score are all critical elements of strong architectures.

Entity Resolution vs Data Deduplication

While data deduplication is related to entity resolution, it is not synonymous.

Deduplication is a question of whether two records are identical. Entity resolution is asking themselves if the records refer to the same real-life entity, and how the business should deal with that connection.

This disparity matters when teams create Master Data Management, Customer 360, supplier mastering, corporate hierarchies or create enterprise data that's AI ready.

Entity Resolution vs Record Linkage and Identity Resolution

These terms are often used interchangeably, but each one has a different emphasis.

Record linkage typically refers to the process of linking records across multiple data sets without a common reliable identifier.

Record matching is the actual comparison step used to determine if two records are a match.

When the entities are people, customers, users, or digital identities, identity resolution is the most common term used.

The most general of these terms is entity resolution. It is capable of wrapping up people, businesses, suppliers, products, providers, assets, locations and so much more.

In real MDM environments, all these techniques are usually combined within the same mastering process.

What Happens After Records Are Matched?

Spotting a match is only the first step. Teams still need to decide what happens to the records once they are linked.

Match and Merge

Records that represent the same entity can be grouped or merged according to clear business rules. The original source records do not have to disappear. They can remain visible and traceable while a mastered version sits on top.

Survivorship

When the contributing records disagree, survivorship rules decide which values appear in the mastered record. One system may be trusted for the legal company name, another for the address, and a third for a classification code.

Golden Record Creation

The end result is often a golden record that holds the enterprise's official, governed view of the entity.

LakeFusion MDM uses entity resolution and match-and-merge to create governed golden records for customers, suppliers, companies, accounts, providers, and other enterprise entities.

Stewardship and Human Review

Not every candidate should be resolved automatically. Cases with low confidence or unclear evidence often need a data steward to look at the records, accept or reject the match, and give feedback. This human review step is especially important when a wrong merge could create operational or compliance problems.

Where Enterprise Entity Resolution Creates Value

The technology delivers the most value when it is tied to a real business need.

Customer 360

Customer information is typically distributed across the CRM, ERP, billing, service, and operational systems. Entity resolution connects the dots to create governed golden records of customers and a more accurate Customer 360 for all teams.

LakeFusion has applied this approach in enterprise Customer 360 projects on Databricks, turning scattered records into trusted master entities.

Supplier Master Data

The same supplier can appear under multiple vendor names, subsidiaries, or regional records. Resolving those records helps procurement and finance teams see which supplier entities are actually the same business and how they fit into larger corporate hierarchies.

Company and Account Resolution

B2B companies frequently need to connect companies, accounts, parent organizations, and subsidiaries. Entity resolution plus hierarchy management gives a clearer picture than simply removing duplicate account names.

Healthcare Identity Resolution

Healthcare systems often hold fragmented patient or provider records across clinical and administrative platforms.

Entity resolution helps create governed identities while keeping the full lineage of the source records.

Resolve Fragmented Enterprise Data with LakeFusion

Duplicate rows aren't the only thing entity resolution can detect. It is all about identifying which customers, suppliers, companies, providers, products, and other records people should consider to be the same real-world entity, and then consolidating those information pieces into data the business can rely on.

LakeFusion is built into its Master Data Management platform on Databricks. The intelligent matching, match and merge, golden record management, survivorship and stewardship, hierarchy management and governance functions can be utilized without the master data being moved into a new disconnected MDM system.

Because LakeFusion runs within the Databricks environment, mastered entities can remain close to the analytics, applications, and AI workloads that need them, while governance stays aligned with the existing lakehouse architecture.

For enterprises already centralizing data on Databricks, the next challenge is not getting more records into the lakehouse.

It is knowing which records belong together.

Explore LakeFusion Master Data Management

Book a Demo

Frequently Asked Questions

What is entity resolution?

Entity resolution is the process of locating information about the same real-time entity when it's housed in two or more systems. When combined with governance rules, matching methods create a more streamlined, more trustworthy customer, supplier, company, product or entity view.

What is the difference between entity resolution and entity matching?

The step where records are compared and it is determined if they match up or not is called Entity Matching. That comparison is part of entity resolution, and the process also covers linking, consolidation, survivorship, governance and sometimes the development of a golden record.

What is record linkage?

Record linkage joins records that live in separate datasets when no shared reliable identifier is available. It is commonly used as one part of a larger entity resolution workflow.

How is entity resolution different from data deduplication?

Data deduplication mainly looks for repeated records. Entity resolution checks whether different records actually stand for the same underlying entity and also decides how those records should be linked, consolidated, and used later.

Can entity resolution support Customer 360?

Yes. Entity resolution can connect the fragmented customer identities that live in CRM, ERP, billing, and operational systems. That connection creates the trusted foundation needed for governed customer golden records and a usable Customer 360.