How it works
Define the acquisition questions, allowed answer values and evidence required to accept each answer. Resolve identity before detailed collection so the system does not confidently research the wrong website or count the same company twice. Ask disqualifying questions first. A system that completes every field regardless of source quality creates apparent coverage at the expense of a trustworthy screen.
Separate collection from review. Collection should retain the page and quote the support verbatim; review should test whether that quote supports the typed answer. Denying fetch tools during the review pass forces it to assess the collected material rather than patching weak evidence invisibly. Contradictions should be settled where possible and unresolved questions kept open. The resulting record should be understandable to the deal team without access to the model’s internal reasoning.
Where it fits
- Repeatable research questions across a substantial candidate population where public evidence can answer meaningful parts of the mandate. A fixed protocol helps expose inconsistent treatment of similar companies.
- Analyst-led programmes that need comparable records and a manageable review queue. Agents can collect broadly while people address question design, difficult cases, relationship context and decisions about further investigation.
Limits and failure modes
- Agents can misread sources, confuse related entities or infer beyond a passage. Separate review helps inspect those failures but does not prove that every accepted answer is correct.
- Public material may be incomplete, stale or inaccessible. Neither a larger model nor more searches can establish private financial facts, owner intentions or contractual terms that the sources do not disclose.
Cost, timing and evidence
- What it costs
- Cost follows the number of companies and questions, source accessibility, retrieval work, model processing and review. Identity ambiguity and contradictory evidence create additional effort. Include analyst attention and source maintenance when comparing workflows; low generation cost does not establish a low cost for an accepted, useful research record.
- First conversation
- Agents can prepare records before contact, but they do not control client approval or owner response. The interval to a first conversation depends on those later stages. Treat research completion, contact readiness and meetings held as separate outcomes, and keep difficult records visible rather than declaring the entire list ready.
- Competitive tension
- Using agents does not make public information exclusive. A buyer may gain a more consistent view of its chosen population, while other buyers can investigate the same companies. The method supports preparation and coverage; any transaction exclusivity still comes from a separate agreement, not the research tool.
- Evidence quality
- Quality depends on source retention, question design and the independence of the review process. A verbatim quotation can still be from the wrong company or fail to answer the criterion. Test entity match, passage relevance and permitted inference separately. A missing answer is more useful than a confident value with no defensible support.
- Scalability
- Typed questions and repeatable collection support broad coverage, subject to source availability and review capacity. Scaling should preserve the same acceptance rules and show the outstanding queue. When criteria change, earlier records may need new research; a completed run is a dated assessment rather than a permanent statement about the company.
Docket’s assessments
These low, medium and high levels describe operating characteristics, not measured results or promises. A high cost burden means greater effort. The other dimensions describe potential access, inspectability and repetition under a well-run process; none establishes acquisition suitability.
| Dimension | Level | Rationale |
|---|---|---|
| Cost burden | medium | Retrieval, processing and analyst review all contribute to total cost. |
| Conversation readiness | low | A supported company record does not establish owner interest. |
| Evidence traceability | high | Retained quotations and separate review can expose the basis of answers. |
| Repeatable coverage | high | A defined protocol can be repeated across a chosen population. |
Review the method as an operating process
Before increasing activity, define what the team needs to learn and which evidence would change its view. A relevant company, a reachable contact, a permitted introduction and a held conversation are separate stages. Record them separately so a busy channel does not appear successful simply because it generates more records. Compare methods against the same mandate and the same review standard.
Use the first set of reviewed records to test the questions themselves. If ownership, service mix or geographic scope is repeatedly unresolved, decide what can be established publicly and what requires a permitted discussion. Leave the latter visibly open. Pause expansion when the team cannot review results or respond properly; research throughput should serve the relationship process rather than outrun it.
Where Docket helps
Docket implements this approach through Triage, Scout and Audit, tailored to each client’s criteria. Clients can bring a list or ask Docket to build one, use self-serve or managed delivery, and export CSV or Excel at any time. Each client’s docket has a separate database and the client owns its data. No agent makes the acquisition decision.
For the distinction between early access and exclusivity, read proprietary deal flow. To apply this method to a mandate, discuss the questions with Docket or inspect the live platform.