Market Dossier: a living view of the market, not a folder of tabs.

The dossier could show a layered market tree. The investor moves companies between segments, adds a missing niche, and saves alternate versions of the map.

Healthcare software
  -> Ambulatory care software
       -> Practice operations
       -> Clinical workflow
       -> Patient acquisition and retention
       -> Physician CRM
            -> Independent practices
            -> Health systems
            -> Specialty-specific products

Segment definitions carry inclusion and exclusion rules, representative customers, geographic scope, and the evidence used to form them.

The useful output is not just a taxonomy. It shows where capital, company creation, customer demand, and consolidation are concentrating, and where there may be whitespace.

Instead of one impressive number, five triangulated views: top-down, bottom-up, value-based, supply-side, and comparable-market. Every result exposes the formula, units, date, geography, currency, source, and editable assumptions. Investors compare "broad narrative TAM," "serviceable market today," and "credible five-year market" rather than collapsing them into one figure.

A scenario workspace could answer: What happens if only independent U.S. cardiology practices are eligible? What if average annual contract value falls 30% but seat penetration doubles? How much of the apparent market is already captive to an incumbent bundle? Which assumptions changed since the investment memo?

Government and international statistics anchor some models but are usually too coarse to be a TAM answer on their own. Their best role is as auditable model inputs, not answers.

More than a logo cloud. Each entity is classified in one of eleven roles: direct competitor, adjacent product moving into the category, incumbent suite that can bundle the feature, substitute including manual work or services, channel or distribution partner, supplier or enabling infrastructure, potential customer, acquirer, investor or strategic backer, newly formed or stealth entrant, and failed, dormant, acquired, or pivoted company.

The system highlights changes rather than forcing a reread of every profile: "moved upmarket," "launched a free tier," "added an FDA-cleared product," "opened three enterprise sales roles." Six visualizations: a two-axis landscape, segment columns with funding and growth overlays, a relationship graph, a timeline of entrants, rounds, launches, exits, and failures, a capability matrix with evidence dates, and geographic maps.

"Unknown" should be a valid cell in every comparison. A sparse, honest matrix beats an AI-filled table of invented claims.

Announced rounds, investors, amount, stage, and stated use of funds. Accelerator cohorts, newly incorporated entities, and stealth emergence. Valuation and round-size distributions where licensed data permits. Time between rounds and follow-on rates. Investor entry, exit, concentration, and syndicate patterns. Debt, grants, government awards, and non-dilutive funding. Acquisitions, divestitures, acqui-hires, bankruptcy, and shutdown signals. Public comparable filings and segment performance.

No single provider should be treated as complete. Funding dates and amounts often disagree, undisclosed rounds are common, and announcements lag legal closings. Show conflicting source claims and let the investor choose a canonical observation.

Open question: coverage, licensing, embedding rights, and economics of each private-market data provider need direct evaluation.

Many of the most interesting signals appear before revenue is publicly visible: category search interest and vocabulary changes, web visits and paid-search activity, mobile downloads and review velocity, marketplace rank, SKU count, stockouts, and retailer expansion, estimated ad volume, customer announcements and integrations, RFPs and awards, exhibitor lists, pricing-page and packaging changes, reseller listings and certification directories.

These weak signals become careful prompts, not conclusions. "Search interest rose" is an observation. "Demand is accelerating" is an inference that needs corroboration.

Open question: traffic and download figures are estimates, small sites are especially noisy, and access to some search-interest data would need to be checked before relying on it.

App-store and marketplace reviews, review sites where permitted, public forums, issue trackers, and support discussions. Review volume, rating distribution, topic frequency, changes after releases, common switching triggers and named alternatives, recurring complaints by segment, and willingness-to-pay clues in pricing discussions and procurement documents.

The output is a "voice of the market" panel with representative cited excerpts, topic trends, and counterexamples.

Sentiment scores alone are too lossy. A negative review from a mismatched consumer segment should not outweigh strong feedback from the target buyer.

Open question: platform access. Own-app review APIs differ from competitive review access, and review-site terms may restrict automated collection or redistribution. Prefer approved APIs, licensed datasets, or user-provided exports, and retain links rather than republishing source content.

Hiring is an imperfect but useful clue to priorities: open-role count and velocity, function, seniority, and location mix, new geographies or regulated-market specialists, sales versus engineering investment, role removals and hiring freezes, executive appointments and departures, and the skills and technologies named in job descriptions.

A useful view explains direction rather than counting jobs: "Competitor A shifted hiring from R&D toward enterprise sales," with links to the roles that support the observation and a warning when the source board may be incomplete.

For technical markets: research-paper volume, topics, citations, and funders; clinical trials, phases, sponsors, and status changes; patents, assignees, claims themes, and citation networks; standards bodies, open-source releases, stars, and dependencies; product documentation, changelogs, and integration catalogs; new manufacturing methods, ingredients, model capabilities, or scientific targets.

Each is one lens. Patents do not equal product quality, repository stars do not equal commercial adoption, publication volume does not equal scientific validity, and trial status does not equal regulatory approval.

Open question: a repository or paper has to be matched confidently to a company or technology before its activity counts as evidence.

Regulation can create a market, change its economics, or eliminate it. Proposed and final rules, guidance, comment periods, and enforcement actions. Approvals, clearances, recalls, adverse events, and shortages. Reimbursement codes and coverage decisions. Privacy, data-localization, labor, and advertising rules. Licensing requirements and state-by-state variation. Tariffs, sanctions, and import/export rules. Standards and certification changes. Lobbying and public comments.

Regulatory data needs domain-aware interpretation and conspicuous disclaimers. An adverse-event report is not proof of cause. A proposed rule is not final law. Link to official records and identify the status and jurisdiction.

Open question: which verticals justify specialized regulatory, scientific, commerce, or procurement sources.

News is most useful when aggregated into a change in narrative: which category terms are gaining or losing prominence, whether coverage is moving from novelty to adoption, backlash, regulation, or consolidation, which analysts, journalists, newsletters, and conferences shape the space, whether coverage is broad or driven by one syndicated story, and how narratives differ by geography and stakeholder group.

Cluster duplicates and syndicated copies. One press release republished 80 times should not look like 80 independent signals.