The methodology behind every line of the dossier.
5 min read
GoSeek builds every dossier from public sources, with a source and date on every claim. Here's where the information comes from, what happens when sources disagree or data doesn't exist, and how the score is calculated.
Each research front reveals one specific thing.
Several agents work in parallel, each on its own front, with its own sources and known limits. Registry records show formal existence: legal name, tax ID, location, activity and shareholders, but not how the company operates. Corporate publications show what the company chose to say about itself. News brings dated facts recorded by third parties, like acquisitions, expansions or leadership changes, without measuring the impact.
Job posts show where money is going now, by role, seniority, city and volume. Financial data, when published, brings revenue, growth and size, and without publication there's no number. Regulation shows what the activity requires and what could block a contract. Reputation shows what customers and press recorded, and an isolated complaint doesn't prove a structural problem. Technology lists the systems that appear in public signals, and not finding a system never becomes a claim that it isn't used.
The result comes from connecting the fronts. An expansion story alone doesn't support a hypothesis, but combined with twenty job posts in the same region and a change in ownership it becomes a dated one. Deep research runs up to 200 searches per account, in a few minutes.
What comes out about the account
Sixteen sections, each one answering a question from the sale
Without a source and date, information stays out of the dossier.
There's one standard. Every claim is tied to its source, the exact excerpt and the date. Each piece of information is an entry in a ledger, with source, collection date and confidence level, and the reference travels with the export to PDF and CRM.
Whatever doesn't meet the standard stays out, with no caveat and no estimate. Confidence matters too: a number published by the company carries more weight than one repeated by others. The quality section closes the dossier by showing how much was verified and where there are contradictions.
How each statement holds up
Without a source, the fact does not go into the dossier
When two sources disagree, both stay visible.
Public sources often disagree: number of stores acquired, profit for a fiscal year, total headcount. In those cases, GoSeek doesn't pick one. Both versions appear in the contradictions block, with source, date and a warning not to state either in the meeting.
Picking a side would make the document cleaner and riskier. A system that decides on its own shifts the risk to whoever talks to the client, without warning. Showing the disagreement hands the decision back to whoever has context and flags the numbers that shouldn't go on a slide.
Portfolio sorted by score
Each company scored against the area's ICP, in code
What wasn't found is stated.
When there's nothing on a dimension, the dossier says so. A company that doesn't publish financials shows up as having no published financials, with no estimated revenue. A system with no public signal shows up as having no evidence, and in MEDDPICC a letter with no backing becomes a gap.
That's why absence doesn't become a low score. A privately held company with no PR team or public job posts leaves little trace, and little trace doesn't mean a bad account. Penalizing companies that publish little would make the score measure internet exposure.
The company dossier
Every sentence with source and date
The score is calculated in code, against your team's ICP.
The score runs from 0 to 100 and is added up outside the model. It reads the ICP criteria set in the platform: segments with industry code groups, size bands by revenue and headcount, roles, triggers and disqualifiers. There are four fixed-weight families, and the screen shows how much each one added or subtracted.
That guarantees two things. The same evidence always produces the same number. And the number depends on who's researching: the same dossier gives different scores to two companies selling different things, because the criterion is each one's ICP.
A score requested from a language model can vary between runs, doesn't show the weight of each factor and can't be challenged item by item. Here, the agents research, extract and date the evidence, and the math is done in code. You can open the breakdown, see which family drove the score and trace it back to the evidence behind it.
How the number is built
You can question the weight, not the math
Public data on companies and professional roles, with an audit trail.
GoSeek uses public information about companies and professional roles, in line with data protection law. A professional role is a function held in an organization and publicly disclosed, such as who's in charge of finance or technology. When personal data is involved, processing has a recorded legal basis. The log records every input, output and change by user, and the dashboard shows the cost of each run.
What GoSeek doesn't do: buy contact lists, use non-public sources, state anything without a source, estimate numbers to fill gaps, hide disagreements between sources or ask a model for a score.
What the research accepts as a source
Material you cannot show will not hold up a proposal
What people ask about methodology.
Is the score a grade the AI assigns?
No. It's calculated in code, from your team's ICP criteria, with four fixed-weight families added up outside the model. The same evidence always produces the same number, and the screen shows what each family contributed.
Where does the information in the dossier come from?
From public sources: registry, the company's website and publications, news, job posts, financial reports when available, regulatory data and reputation. Nothing comes from purchased lists.
Can I see the source of a specific line?
Yes. Every claim opens the reference behind it, with the exact excerpt and date, in the PDF and CRM too.
What happens when two sources disagree?
Both versions appear in the contradictions block, each with its date, and the point is flagged as disputed, with a warning not to state either in the meeting.
What data does GoSeek use in its research?
Public information about companies and professional roles, with an audit trail, in line with data protection law and with a legal basis when personal data is involved. We don't buy contact lists.
Other pages from GoSeek.
Lead, prospect and account.
Lead, prospect, account and opportunity are not synonyms. Understand the difference and why the account, and not the contact, is the unit of work in B2B.
AI in the sales process.
Where AI fits into each phase of the B2B sales process, what it should not decide, the mistakes that sink the project and how to measure the gain.
Marketing and sales with AI.
Marketing delivers volume, sales complains about quality. How to write a shared definition of a qualified lead and what AI changes at that boundary.
AI agents for prospecting.
AI agents for prospecting read the account in public sources and deliver the company dossier with a score, dated triggers and who to approach first.
AI SDR.
AI SDR has become the name of three different things. What each one does, what the advertised numbers hide and where account research comes in.
Prepare the sales meeting.
How to prepare a B2B sales meeting: what to study about the company, the meeting flow and the questions that only work after account research.
B2B data enrichment.
B2B data enrichment beyond industry code and size: what the company announced, who decides and what changed, with a source and date on every field.
B2B buying signals.
B2B buying signals are dated events that change the chance of closing now. See the catalog of sales triggers and how to prioritize accounts by timing.
Map the decision makers.
Stakeholder mapping in B2B sales: how to identify a company's decision makers in public sources, separate power from interest and keep the map alive.
Sales battlecard.
Sales battlecard: what goes into it, why most of them age inside a slide and how to build one per account, with competition, objections and dated evidence.
B2B prospecting tools.
B2B prospecting tools solve different problems. See the 5 categories, who is in each one and how to choose based on the stage that is stuck.
Your next meeting can start with the right information.
Leave your email and we'll show you a dossier on a company in your industry.