Loading...
Signal scores companies on six dimensions of AI maturity using up to eight public data sources. Every report shows exactly which sources responded. No surveys, no self-reporting, no marketing decks. Just what the data shows.
Every Signal report starts with raw data collection across up to eight public sources. We pull job postings, GitHub activity, SEC filings (for public companies), funding data, press coverage, leadership profiles, Glassdoor signals, and company websites — all in parallel, each on a strict timeout. Private companies are scored with the same rigor using the sources available to them.
The raw data is drafted into dimension scores by our AI scoring engine, then a deterministic post-processor — plain code, not a model — recomputes the overall score, the Narrative Gap, and the verdict, and applies a named-rule catalog. When too few sources respond, the score is capped; when data is missing for a dimension, it scores null with low confidence. We never fabricate signal.
The flagship metric. The Narrative Gap is the delta between a company's Public Narrative score and its Product Embedding score. A company that talks about AI more than it ships AI will show a large positive gap.
Gap > 20 pointstriggers the "Marketing-Led" verdict override — regardless of the overall score. This is the theater detector.
A gap of 10 or less is healthy. Between 10–20 is worth watching. Above 20 means the company is managing perception more carefully than capability.
A worked example, end to end — the same pipeline every report goes through. Watch how a company with a respectable score still lands a “Marketing-Led” verdict.
Up to eight public sources are queried in parallel, each on a strict timeout. A source that fails returns nothing — it is never guessed. Here, Crunchbase has no data for the company, so seven of eight sources feed the score, and the report says so.
The scoring engine drafts a 0–100 score for each dimension, with evidence citations and a confidence level. Investment Signals has nothing to score — so it gets null, not a guess.
The missing dimension's weight is redistributed across the five that scored — deterministic arithmetic, not model judgment. Weighted average: 60. If too many sources had failed, the score would be capped, or the verdict declared undetermined.
A catalog of named, deterministic rules — code, not the model — checks the draft: theater detection, hiring theatre, GitHub ghost towns, single-source caps. Every rule that fires is listed in the report, with its reason. Here, the Narrative Gap crosses the line.
A respectable 60 — but the verdict is Marketing-Led, because the gap rule outranks the average. That's the point of the system: the number you brag about is the one we check hardest. Every threshold in this example is the real one from the scoring code.
7 data sources queried in parallel. Each fetcher has a 10s timeout and returns null on failure.
Raw data is structured into a typed schema. Missing sources are marked, not faked.
AI drafts each dimension 0–100 with evidence citations and confidence levels. Weights renormalize over the dimensions that actually scored.
A deterministic rule catalog — not the model — recomputes the overall score, applies caps and floors, and settles the verdict. Every fired rule is cited in the report.
Measures the volume, specificity, and seniority of AI-related job postings relative to industry peers. Generic 'AI experience preferred' listings score low; specific roles like 'Staff ML Engineer — Retrieval Systems' score high.
Evaluates whether AI strategy is backed by genuine executive commitment. Looks for dedicated AI leadership roles, board expertise, and whether claims trace to organic capability or acquisitions.
Analyzes actual technical infrastructure for AI. Distinguishes between companies running production ML pipelines and those simply mentioning AI in marketing materials. Tech references on the company website corroborate stack depth only when confirmed by GitHub or job descriptions.
Tracks financial commitment to AI through acquisitions, R&D allocation, strategic partnerships, and disclosed investment figures. For private companies, Crunchbase funding data and press coverage carry primary weight. Words are cheap — money is signal.
Quantifies the volume and intensity of a company's public AI messaging — including their own website marketing claims. A high narrative score isn't inherently bad — but when it outpaces product reality, the gap becomes the story.
Measures whether AI features are actually shipping in products users touch. We scrape company product pages for real AI feature evidence and cross-reference against GitHub activity and hiring patterns.
Repository activity, contributor patterns, commit history, language distribution
Code doesn't lie. Active ML repositories with meaningful commit patterns indicate real engineering investment.
Active listings from LinkedIn, company career pages
Hiring intent is a leading indicator. Companies building AI capabilities need people to build them.
10-K, 10-Q, 8-K filings, earnings call transcripts (public companies only)
Regulatory filings carry legal weight — companies are more careful about claims made to the SEC. For private companies, this source is gracefully skipped and other signals are weighted more heavily.
Funding rounds, acquisitions, investor profiles
Follow the money. Acquisitions and investment patterns reveal strategic priorities.
Press releases, media coverage, analyst reports
Measures the narrative layer — what companies want the market to believe about their AI story.
Executive profiles, board composition, organizational structure, team/about pages
Real AI commitment shows up in org charts. Dedicated AI leadership signals long-term strategy.
Homepage, product pages, AI-related content, careers and team pages
First-party claims are the baseline for narrative gap analysis. What a company says on its own site is compared against hard evidence from every other source.
Employee reviews mentioning AI tooling, ML teams, and engineering culture
Employees describe the company they actually work at. Silence about AI on the inside — while the outside is loud — is itself a signal, and one of our named rules.
Score 78+ with a Narrative Gap under 12, and at least five dimensions scored. The company is genuinely building and shipping AI capabilities. Evidence backs the claims.
Score 55+ with a Narrative Gap of 20 or less. Real effort underway, but gaps remain between ambition and execution.
Narrative Gap above 20 — regardless of overall score — or a score below 55. The story runs ahead of the substance. More theater than transformation.
Too few data sources responded to score responsibly. We can't score what we can't see — and we won't guess.
Signal is a point-in-time snapshot, not a continuous monitor (unless you're on Pro). Reports reflect data available at scan time and are reused for up to 30 days before a fresh scan re-scores the company.
We rely on public data only. Companies with strong private AI efforts and poor public signaling will score lower than their true capability — that's a feature, not a bug. If it's not public, we can't score it.
The scorer is an LLM. Like all LLMs, it can hallucinate or misjudge nuance. Every dimension includes a confidence level. Low-confidence scores should be treated as directional, not definitive.
Analyze any company and see seven sources scored across six dimensions in 60 seconds.
Analyze a Company — Free →