Pick two to four entries from the Israeli Startup Database and read them side by side: entity kind, status, exposure, sector context, research signals, and how fresh each research record actually is. The matrix keeps different entity kinds visibly different instead of pretending they compete on one scale.
Database records as of 2026-06-01 · profiles remain the canonical record
No companies selected yet. Search above or load the example comparison.
What kind of entry each column is — and which evidence standard applies before you read anything across columns.
Published database attributes side by side. Every cell traces back to the linked profile page, which remains the canonical record.
Scores are research signals for prioritizing further reading — not investment rankings, quality grades, or return forecasts. Read them within a single company profile only, and never across different entity kinds.
The database deliberately holds more than private startups: public companies, acquired assets, defense primes, government-owned companies, funds, and strategic references all matter to the research picture. Each kind carries its own evidence standard, so the matrix pins an evidence note to every column and raises a warning whenever a selection mixes kinds.
Every column shows when its profile was last updated and flags records older than 12 months as stale. Staleness measures the research record, not the company — an active company with a stale profile simply has not been re-verified recently, and its attributes deserve extra suspicion.
Use the matrix to decide which profiles to open next, then continue in the Startup Database, the linked Dependency Atlas investment cases, and the diligence checklists before treating any cell as verified.
A comparison matrix is the most tempting format in research and the easiest to misuse. Two companies placed in adjacent columns look instantly comparable, even when the columns describe entirely different kinds of things. That is why this tool starts from the entity-context strip rather than the numbers: before any attribute is read across a row, the reader should know whether the columns are private startups, public companies, acquired assets, funds, or ecosystem references. A fund is evaluated on manager track record and fee and carry terms. A public company is evaluated on audited filings and liquidity. A private startup is evaluated on customer traction and technical claims that usually cannot be settled from public sources at all. When the selected entries span different kinds, the matrix shows a warning and keeps showing it until the mix changes — because no layout choice can make those evidence standards interchangeable.
The second discipline the matrix enforces is freshness. Every column carries the date its profile was last updated in the Claw & Talon database, and any record older than twelve months is flagged as stale. The flag is deliberately blunt. Company status, funding stage, headcount, and product claims all decay at different speeds, and a single date cannot say which cell has expired. What it can say is how much independent re-verification the reader owes before quoting anything. A fresh profile means the research record was recently checked against public sources; a stale one means the burden of proof has shifted back to the reader entirely.
Score signals sit in their own panel, rendered within each company only, and that separation is intentional. The research scores in the database — potential, technology, market, team, dual-use, and strategic alignment — are signals for prioritizing further reading, produced under the site's published editorial method. They are not investment rankings, quality grades, or return forecasts. The matrix therefore refuses to highlight a "winner" across columns, refuses to sort companies by score, and suppresses scores entirely for entity kinds where product-style scoring is meaningless, such as funds, ecosystem references, and unverified records. If two adjacent bars invite a cross-column conclusion, the correct move is to open both profiles and check what evidence sits underneath each number.
It is equally important to say what this app does not do. It does not rank the selected companies, estimate valuation, or assess suitability for any investor. It does not know whether a company is raising, available, or willing to engage. A match between entries — same sector, same stage group, same Atlas theme — is a research adjacency, not an endorsement of either company or a claim that they compete head to head. The matrix also cannot see anything the database cannot see: customer references, revenue quality, cap tables, and procurement pipelines live outside public sources and outside this tool.
Finally, the profile pages remain the canonical record for every entry shown here. The matrix intentionally displays a compressed slice of each record so that columns stay readable; the full description, source list, verification notes, and investor-lens questions live on the linked profile. A sound session ends with those pages open — and with the diligence checklist for the relevant entity kind, which every column links to directly.
Those boundaries are the point of the tool. A comparison that refuses to rank stays honest about what public research can and cannot establish. Questions past that line — pricing, allocation, access, timing, suitability — require direct diligence and, where appropriate, legal, tax, regulatory, or technical advice.
Claw & Talon materials are for general informational and research purposes only. Side-by-side attributes, research scores, freshness flags, and entity-type notes do not constitute investment advice, a recommendation, ranking, suitability analysis, offer, solicitation, endorsement, or an indication that any investment is available.
