Kscope · competitive strategy · 2026-10-08

The Scoring Layer

The data moat is built. The value we haven't captured isn't more data — it's the model on top.

We scored 156 companies in the business-information market on the data, not the screen. The finding is blunt: across the market, the data half is overwhelmingly ours or free — filings, ownership, insiders, transcripts, people, pay, the private graph from ~100,000 news/RSS/crawl feeds in 16 languages, all joined on one entity graph. What we don't yet ship, company after company, is the model / score / opinion on that corpus.

Now

Data we already extract & join

Fast

A model on inputs we hold or can crawl

Slice

We take the core; an input stays licensed

Never

A contributed book / panel / wire

Of the 67 reachable (now + fast), a large share is reachable-but-unbuilt for one reason: the product is a model, score, or opinion on data we already hold — not a dataset we'd have to buy.

Why this is the next phase, and why it's high-leverage.
(1) The inputs are already in house — these are models on data we extract + join, not new ingests. (2) Our models are a superset, not a clone, because of the one graph: a 13F filer, a 13D position, a Form 4 person, a 990 trustee, a proxy officer and an agreement counterparty all sit on the same company / person id. Every competitor keeps those segments apart.

The surface catalog

Ranked by value × how much of the input we already hold.

Tier A

Start here

Inputs 100% in house, clear demand, modest build.

Surface	Market example	Our input (held)
13F scoring + clone backtester + top-held indexseeded	WhaleWisdom ($90–150/qtr)	77M+ 13F rows · 11,646 managers · our manager score · total-return prices
Insider scoring + auto-brief	VerityData (InsiderScore)	Form 3/4/5 classified · 10b5-1 · buybacks · grants · footnotes
M&A league tables	Dealogic / ION	announced M&A · advisers · both sides' holders
News factors / signals	RavenPack · Accern	the RSS/GDELT/filing corpus + signals extraction

Tier B

Flagship-grade

Higher value, slightly larger build.

Surface	Market example	Our input (held)
Credit scoring / opinion flagship	Moody's · S&P Global Ratings	filings · agreement clauses · credit/debt tables · news
Fundamental rating / fair-value / moat	Morningstar	xbrl fundamentals · the N-PORT fund book
Private-market prediction + taxonomy	CB Insights · PitchBook	the funding / M&A / people graph · Form D/C
Governance / activism analytics	Diligent · Equilar	DEF 14A · pay · the 13D activism trail · boards

Tier C

Adjacent / later

Needs a data build first, or is a separate play.

ESG scoring — a data gap before a score.
Transcript analytics — sentiment / Q&A / guidance tone on ~264K transcripts.
Foundation / donor scoring — the healthcare-philanthropy play (Candid / GOBEL).
Fundamentals-extraction KPIs — adjusted/segment KPIs from 8-K exhibits (Daloopa-style).

The flagship: a credit score / opinion. Highest value, the inputs are already held (filings, agreement clauses, the debt stack, news), and it's the single best showcase of the superset edge — a credit view that also knows the insiders selling, the activist circling, the bondholders, and the board. We publish our view, never a copy of a rating letter.

What stays theirs

Inputs we can't model our way to — don't chase them.

Live / intraday prices · estimates & consensus (I/B/E/S, StreetAccount) · the wire & news archive (Dow Jones/Factiva, the Benzinga squawk) · contributed books typed in by participants (PitchBook, BoardEx, Equilar survey data) · LP cashflows & fund performance (Preqin) · and the rating letter itself (Moody's/S&P assign it). The rule: we publish our own model or view, never a copy of their licensed output.

Sequencing

Live start — the 13F scoring cluster (backtester, index, manager scoring). WhaleWisdom-shaped; inputs 100% in house; already seeded.
Flagship — the credit score / opinion. The clearest demonstration of "a model on data we hold that's a superset because of the graph."
Cheap wins — insider scoring, M&A league tables, a news factor. All inputs fully in house.

Method. Scored on the data layer, not the UI, across 156 companies in the business-information market (kscope-vs-market), with deep-dives on Candid, Law Insider, BamSEC and WhaleWisdom. A row counts as reachable when the raw material is free or the product is a model on data we already hold. Companion to the business-information landscape. — Kscope, 2026-10-08.