Entity Resolution for Prospect Research: Connecting Legal Names, DBAs, Licenses, and Locations

5 min read · SiftSpot Insights

A public record is only as useful as the identity attached to it.

That sounds obvious until a business appears under four names.

The storefront may use a memorable brand. The food-service license may list an LLC. The alcohol license may show an owner or primary name plus a DBA. The corporate registry may list managers under the legal entity. The website may identify a hospitality group that operates several locations. A mailing address may belong to an accountant or management office rather than the venue.

If those records are merged too aggressively, the research becomes wrong. If they are never connected, the same business appears repeatedly as separate prospects.

Entity resolution is the middle discipline: connect records when the evidence supports the connection, and preserve separation when it does not.

Start with a field model, not a company-name column

A single “Business Name” field is not enough.

At minimum, keep separate fields for:

• legal entity name;
• DBA or fictitious name;
• public-facing brand;
• licensee or owner name when the source uses a separate field;
• operating address;
• mailing address;
• license number;
• corporate document number;
• relevant officers, managers, or principals;
• source and source date.

This structure mirrors the way public systems actually represent businesses.

Florida’s Sunbiz search tools reinforce the point. The Division of Corporations allows searches not only by entity name but by officer or registered-agent name, document number, ZIP code, street address, and other identifiers. Corporate detail records can then expose current officers or managers and filed documents. [S1] [S2]

DBPR hospitality records add another layer by separating licensee name, business or location name, address, and license number. [S3] [S4] [S5]

Use stable identifiers first

Names are useful. Identifiers are stronger.

A license number, corporate document number, exact operating address, or authoritative registration number can connect records more reliably than a similar business name.

A reasonable matching order is:

1. exact authoritative identifier;

2. exact location plus compatible legal/DBA name;

3. legal entity plus supporting address or principal;

4. DBA plus multiple corroborating fields;

5. name similarity alone, which should remain low confidence.

The objective is not to produce a match at all costs. It is to produce a confidence level that another reviewer can understand.

Keep legal operator and brand separate

A restaurant brand can outlive the entity that operates it. A hospitality group can operate multiple concepts through multiple entities. An owner change can leave the sign on the building unchanged.

That is why the question “Is this the same business?” may have more than one answer.

It may be the same brand but a different legal operator.

It may be the same legal entity at a different location.

It may be the same management group operating a different concept.

It may be a completely unrelated business with a similar name.

Do not force these relationships into a single yes/no flag. Store the relationship type.

Example:

• Brand: Harbor House
• Legal operator: Harbor House Dining LLC
• Location: 100 Example Avenue
• Relationship: legal operator of this branded location

If a later record shows Harbor House Hospitality LLC at the same address, the new record should not silently overwrite the old entity. It should trigger a possible-ownership-change review.

Treat addresses as evidence, not proof

Addresses are powerful matching fields because regulated licenses are often tied to physical locations. They are also imperfect.

One address can host different businesses over time. A shopping center can generate near-identical addresses with suite differences. Corporate filings may use a registered-agent or mailing address unrelated to operations. A hotel can contain separate restaurants or bars with their own records.

Normalize addresses for comparison, but preserve the original source text. If “Ste 200” is standardized to “Suite 200,” retain the source value somewhere in the record so a reviewer can see what the agency actually published.

Use principals carefully

Officer, manager, member, registered-agent, and licensee names can help connect records, but their roles matter.

A registered agent is not automatically the business owner. An attorney or professional registered-agent company can appear across many unrelated entities. A corporate officer may have changed since the last filing. A manager listed in one system may not be the right commercial contact.

The right use of a person field is corroboration: Does this name strengthen the match between two business records? It should not automatically become the outreach contact.

Ownership and legal-entity changes deserve special handling

Some government registration systems explicitly treat ownership and legal-entity changes differently from ordinary profile updates.

The Florida Department of Revenue, for example, instructs businesses to notify the department about certain name, address, closure, or sale changes, while stating that a new registration is required for changes such as a new legal entity or ownership. [S9]

For prospect research, the lesson is not to infer tax status. The lesson is that “same location, new entity” can represent a materially different administrative event from “same entity, updated mailing address.”

Your entity model should be capable of preserving that difference.

Create match confidence rules

A simple confidence scheme keeps the process reviewable.

High confidence

• exact authoritative identifier match; or
• exact operating address plus matching legal entity/DBA and a second corroborating field.

Medium confidence

• strong name and address match but one key identifier is missing; or
• matching legal entity and principal with a plausible but not fully confirmed location relationship.

Low confidence

• name similarity without stable identifiers;
• shared registered agent only;
• address similarity with conflicting names;
• brand association inferred only from a third-party directory.

Low-confidence records can still be useful. They simply belong in an identity-resolution queue, not an outreach queue.

Deduplicate at the organization level

Commercial prospecting systems often deduplicate people but forget to deduplicate organizations.

If three officers at the same independent agency are discovered from different sources, that should not automatically create three prospects. If a restaurant group operates several venues, the correct outreach strategy may be location-specific, group-level, or both—but that is a deliberate business decision, not something a matching algorithm should invent.

Keep an organization key that can connect contacts, locations, and records without erasing their differences.

Preserve history instead of overwriting

Entity resolution is temporal.

The record that was correct last month may be superseded by a new filing. A license can change status. A manager can change. A DBA can be added. A location can close.

Do not simply overwrite yesterday’s values with today’s. Preserve the change history:

• previous value;
• current value;
• source;
• effective or filing date when available;
• date observed.

That history is often the most useful business-change intelligence in the system.

The standard for a defensible match

A reviewer should be able to answer three questions:

1. Which fields caused these records to be connected?

2. Which source supports each field?

3. What uncertainty remains?

If those answers are visible, entity resolution becomes a trustworthy part of the research process. If they are hidden, the system eventually accumulates false matches that look authoritative because they have been copied forward many times.

Good prospect intelligence begins with identity discipline. Before asking whether a business is worth contacting, make sure it is the business you think it is.