Your proprietary data source. No vendor sells this.
Because it is built from what only you have. Every data subscription sells the same database to you and to everyone else. Kruncher builds a data source that exists nowhere else.
Three things combined, and only one of them is for sale
A subscription gives you the second. The first and the third are yours, and putting all three in one record is what makes the result proprietary.
The material that never leaves your firm
CIMs, teasers, banker decks, management presentations, meeting notes, call transcripts, prior diligence, and the judgment your team has accumulated across every deal it has looked at and passed on. None of it exists in any database anyone can buy.
- Documents and data rooms
- Notes, transcripts and email
- Decisions taken: liked, passed, invested
Refreshed continuously, not quarterly
More than twenty premium sources, the public web and country level registries, reconciled into the same record rather than sitting in twenty browser tabs. Where a snapshot database gives you the last quarter, this gives you the last week.
- 20+ premium databases and registries
- The public web, multilingual
- Reconciled with provenance, not appended
Your categories, not a vendor's tags
Every company classified the way your firm classifies companies, and scored against the criteria your firm actually uses. Two firms running the same company through Kruncher get two different answers, because the lens is theirs.
- Your sectors, stages and segments
- Up to 500 scoring parameters, weighted by you
- Learned from what your team marks up and down
The combination is the thing, not any one source
Your documents alone are incomplete. External data alone is the same data your competitors bought. Reconciled into one record, each one makes the other worth more.
Every value keeps the layer it came from, so a reader can always see whether a figure is what the company claimed, what the public record shows, what Kruncher estimated, or what your own team wrote down.
Sources in at the bottom, your own dataset out at the top
Your sources, your diligence playbook and your decisions feed the ingestion layer. The knowledge graph connects what comes out. Workflows, analytics and your own agents read from it.
Companies, people and investors, and everything between them
Kruncher maintains an entity resolved graph over millions of private companies. It is not a list with tags. Every edge has a type, a direction and a source.
What surrounds a company is the part nobody sells
A snapshot database answers what a company is. The graph answers what surrounds it: who backs it, who competes with it, who sells to it, who buys from it, who founded it and who runs it. An analyst sees the syndicate, the customer base, the vendor dependencies and the competitive set without running twenty searches.
- Founders and executives tracked across every company they have touched
- Competitors, vendors, customers and partners, typed and directional
- Investors, syndicates and the funds that hold each position
- Relationship claims resolve to a source, which is where generic tools hallucinate
- Your own companies and contacts join the same graph, and stay yours
You define the output, and you define what reaches you
The report is a template your firm owns. The signals are rules your firm writes. Neither is something a vendor decides on your behalf.
A report in your template
Choose the sections, add the ones only your firm asks for, reorder them, and score each against your own criteria. The same template then runs on every company you analyze, so an output arrives already in the shape your investment committee reads.
Every section is scored against your own criteria and written into your own template, so two firms running the same company get two different reports.
And it is proactive
You define the signals that matter to you: a competitor launch, a target raising, a key hire, an expansion into your geography. When something relevant happens, it is sent to you. You do not query it. It tells you.
You do not query it. When something you said mattered happens, it reaches the person who owns the answer.
Built for private markets, which is why it is more precise
A general purpose extraction tool reads a document. It does not know what the document is, what the numbers mean, or whether the answer it produced is plausible. Six places where that difference shows up.
Funds, vehicles, feeders, AIVs, LPs, GPs, commitments, capital accounts, rounds, cap tables, marks and holdings are first class objects, not custom fields bolted onto a generic CRM. A generic tool has no idea what a sleeve is.
A CIM reads differently from a teaser, a capital account statement differently from a quarterly report, an LPA differently from a side letter. The engine classifies the document before it reads a figure, because what a number means depends on where it was printed.
Gross versus net of accrued carry. Reference date versus file date. ARR versus revenue versus bookings. Fully diluted versus outstanding. TVPI, DPI and MOIC computed the way the industry computes them. These are the places a generic extractor quietly gets it wrong.
Top down NAV computed against bottom up NAV. Implied multiples sanity checked. Bridges that have to reconcile. A general purpose model has no view on whether a number is plausible. A private markets engine does.
Public comparables are easy and everybody has them. Knowing which private companies are genuinely comparable, who competes with whom and who sells to whom, is what the graph is for, and it is built from 200,000 company analyses rather than scraped tags.
600 configurable signals across people, liquidity, M&A and business deals, defined as the moments that move a private market decision rather than as generic news alerts. You add your own on top.
An information advantage your competitors cannot buy, because it is built on your data and your judgment.
It is not pooled, not resold and not used to train shared models. It can run inside your own cloud, against models your risk function has already approved, and it is reachable over the API and the MCP server like everything else. How that is kept yours.
Bring the data only you have.
Connect one document store and one inbox on the call, and see what the record looks like when your own material sits beside twenty external sources.