A newly incorporated company appears on Companies House. The director’s name is available. The obvious next step seems simple: find that person on LinkedIn.
In practice, this is one of the hardest parts of turning public company data into usable sales intelligence.
The problem is not search. It is identity.
Why a name match is not enough
A name can produce multiple plausible LinkedIn profiles. Geography can be inconsistent. People move. Their profile may describe a profession that does not obviously match the SIC code selected for a new company. The company may be so new that it has no website, LinkedIn page or meaningful search footprint.
A system that simply selects the closest-looking result will achieve an impressive completion rate and quietly contaminate the dataset.
Our objective is different: establish whether the available evidence supports the conclusion that the online profile and the Companies House individual are the same person.
Start with the authoritative identity anchors
The Companies House record provides the starting identity context. Depending on the available record, useful anchors can include the person’s name, appointment relationship, month and year of birth where publicly supplied, nationality, country of residence, service address context and relationships to other companies or persons with significant control.
These are not all equally useful, and they should not be treated as a simplistic points checklist. Their value depends on the case.
Build a candidate, then try to disprove it
Finding a plausible LinkedIn profile is only the candidate-generation stage.
The next stage is corroboration. Does the person’s location make sense? Does their professional history connect to the new company’s activity? Are there other public sources linking the person to the same geography, sector or business? Does another directorship provide a bridge? Is the profile supported by a company website, professional biography, event page, publication or other independent evidence?
Equally important: is there contradictory evidence? A candidate with the same name who is clearly based in another country, works in an unrelated field and has no supporting connection should not be rescued simply because the name is distinctive.
The absence of a new company on LinkedIn is not a contradiction
This point matters particularly for new incorporations.
If a company was formed hours or days ago, it may not yet appear on the director’s LinkedIn profile. It may not have a website. Search engines may know almost nothing about it.
That absence is neutral. The verification process should look for independent anchors around the person rather than penalising a candidate because the brand-new entity has not yet developed an online footprint.
Confidence should control release
We use confidence as an operational control, not decoration.
A high-confidence result can be released into the next stage of the workflow. A probable result can be held for further research or review. An unresolved result remains unresolved.
The exact threshold can be configured around the use case, but the principle is consistent: the next action should reflect the strength of the evidence.
This is particularly important when enrichment follows. Once a LinkedIn identity is released, downstream providers may return emails, mobile numbers and additional attributes. If the identity is wrong, technically accurate enrichment can still attach to the wrong human being.
What good evidence looks like
Strong matches often come from combinations rather than a single magic field.
A distinctive name plus matching geography and a professional niche consistent with the new company may be persuasive. A common name may require a second directorship, employer history or independent biography. A geography discrepancy may be explainable by relocation, but it should trigger more research rather than being ignored.
We also record the reasoning behind a decision. That makes the system easier to audit and improve because a reviewer can see which positive evidence, contradictions and uncertainty flags led to the outcome.
AI helps, but it needs boundaries
AI research agents are well suited to collecting and interpreting messy web evidence. They can compare names, locations, careers, company descriptions and search results far faster than a person manually opening every tab.
But an unconstrained model is also capable of turning weak clues into a confident narrative.
That is why the workflow matters: explicit research steps, defined evidence expectations, contradiction handling, confidence rules and a valid unresolved outcome.
The agent should be allowed to say, in effect, “I found a likely person, but I cannot support the match strongly enough.”
Why this matters commercially
Identity resolution is invisible when it works. The salesperson simply receives the right person.
But it changes everything downstream: enrichment accuracy, relevance of messaging, compliance processes, brand experience and the amount of time salespeople waste correcting bad records.
This is why we describe the New Incorporation workflow as identity first, enrichment second.
Companies House gives us an excellent public-data foundation. The engineering challenge is turning that record into a sufficiently verified human identity before the sales process acts on it.
Ask Sentra to show you sample resolved incorporation data.
Source note: Companies House makes public company information available through its register and API. The identity-resolution methodology described here is Sentra’s own workflow and is not a Companies House service or endorsement.