How Foxhole scores companies
No black box. Here's exactly what we measure and why.
01Signal taxonomy
Foxhole watches fourteen signal types across the public surface area of every company on the radar. Each signal carries a weight, a detection method, and a stated reason for inclusion. Active signals (those producing scores today) are listed first; signals in the constraint set ready to activate as collectors ship are listed below.
Engineering Acceleration
Tier 1 · high weightWhat it measures
A sustained spike in commit cadence across a company's public repositories compared to its own 30-day baseline.
How we detect it
We pull commit metadata from public GitHub organisations on a daily cycle and compare current-window throughput to baseline. Ratios above 1.7x register as a signal.
Why it matters
Commit velocity doubles three to six weeks before fundraise announcements and major enterprise releases. The pattern shows up before any wire hits.
Executive Function Hire
Tier 1 · high weightWhat it measures
A first-time senior leadership hire that signals a transition from founder-led to function-led execution: VP Sales, Head of Operations, CFO, COO, Head of Customer Success.
How we detect it
Daily scan of company About / Team / Careers pages, cross-referenced with public LinkedIn role announcements and press releases.
Why it matters
The executive function hire is the canonical "we just outgrew our seed motion" signal. Bootstrapped companies making it tend to be preparing for either enterprise scale or a first institutional round.
Enterprise Hiring
High weightWhat it measures
Roles that only exist at companies entering enterprise sales motion: VP Sales, Enterprise AE, Solutions Engineer, Customer Success Manager, RevOps lead.
How we detect it
Daily ingest of public job postings across major boards (LinkedIn Jobs, AngelList, Wellfound, company career pages). Classifier groups roles into enterprise-motion buckets.
Why it matters
Hiring an enterprise sales motion changes a company's runway math. If a bootstrapped company starts hiring AEs without first raising, the next move is usually one of two things: a contract pipeline already exists, or a round is imminent.
SOC2 / Compliance Hiring
High weightWhat it measures
A first-time security, compliance, or controls hire: SOC2 Lead, Compliance Officer, Security Engineer (compliance focus), or DPO / Data Protection Officer roles.
How we detect it
Pattern-match against the same job posting feed used for Enterprise Hiring, with a separate keyword set tuned to security and compliance language.
Why it matters
Enterprise prospects ask for SOC2 before they sign. A company hiring its first compliance lead is preparing for contracts that require it. This is operational scaling without the press release.
Contributor Growth
Medium weightWhat it measures
An expansion in the unique-contributor count across a company's public repositories, especially when the new contributors are full-time team members rather than community PRs.
How we detect it
Track unique commit authors across public organisations over rolling 30-day windows. Filter for committers whose contribution velocity suggests employment rather than drive-by community contribution.
Why it matters
Quietly growing the contributor pool is a leading indicator of team expansion that often pre-dates any public hiring announcement.
Star Acceleration
Medium weightWhat it measures
An acceleration in GitHub stars on a company's main public repositories, normalised against the repository's own historical growth rate.
How we detect it
Daily star-count delta against the trailing baseline. Acceleration ratios above 2x register as a signal. We weight new-star sources for quality — a star from a recognised practitioner counts more than a star from a brand-new account.
Why it matters
Star acceleration is a community-pull signal. It tells you the practitioner audience is starting to notice. For open-source-first companies, it precedes enterprise pull-through.
Serial Bootstrapped Builder
Tier 1 · high weightWhat it measures
A founder track record of shipping multiple public products independently — not first-product founders, but operators on their third or fourth shipped thing.
How we detect it
Cross-reference founder accounts against ProductHunt maker histories, GitHub commit graphs across multiple repositories, and public domain registration patterns.
Why it matters
The pattern predicts which lean teams will actually scale. A founder who already shipped three products knows the parts of building that don't get romanticised. They tend to underpromise and outbuild.
Bootstrapped Milestone Tweet
High weightWhat it measures
A public revenue, customer, or growth milestone announcement from a verified founder of a bootstrapped company.
How we detect it
NLP classification of public social posts from verified founder accounts. Filter for milestone language ($X ARR, Y customers, Z% growth, profitable, sustainable, etc.) tied to a known bootstrapped entity.
Why it matters
Bootstrapped founders only announce milestones they want enterprise prospects (or future investors) to know about. The post itself is a deliberate signal of inflection.
Signal types ready to activate
The following signal types are in the scoring constraint set and ready to activate as their collectors ship. They appear in the schema today; live data appears in scoring as each collector goes online.
Cross-Platform Pain Mentions
Tier 1 · high weightWhat it measures
Velocity of unmet-need conversations on public forums, GitHub issues, and product community threads — the categories of problems users are looking to solve before the companies that solve them exist.
How we detect it
Cross-platform crawl of public discussion surfaces, NLP-tagged for pain-signal language and grouped into emerging categories. Acceleration ratios above 1.5x flag the category.
Why it matters
The most asymmetric calls in venture come from spotting a category in motion before any company has scaled into it. This signal is the leading edge of that.
Emerging Vendor Recommendation
Tier 1 · high weightWhat it measures
Practitioners on public threads recommending a specific vendor as the "new" or "better" alternative to incumbents in a category.
How we detect it
Pattern-match against public posts containing recommendation language paired with a vendor name, where the vendor has below-threshold incumbency in the category.
Why it matters
Word-of-mouth from practitioners is the most expensive signal to fake. When a vendor is genuinely being recommended in their category, scale follows.
Tech Stack Migration
Tier 1 · high weightWhat it measures
A meaningful change in the infrastructure layer beneath a company's product — database, observability stack, deployment platform, identity layer.
How we detect it
Pattern-match against public technical blog posts, conference talks, public job postings citing specific stacks, and public BOM disclosures.
Why it matters
You don't migrate stacks until the cost of the migration is less than the cost of not migrating. The signal tells you a company has reached the scale where infrastructure choices matter more than founder taste.
Operational Scaling Without Capital
High weightWhat it measures
A bootstrapped company with no recent fundraise that is nonetheless hiring across multiple operational functions in parallel.
How we detect it
Composite: funding status from public registries combined with the hiring velocity feed from Enterprise Hiring and SOC2 Hiring signals.
Why it matters
This is the operational definition of "bootstrapped company entering enterprise scale." The signature signal of Foxhole's whole thesis.
Founder Substack Growth
Medium weightWhat it measures
Audience and engagement growth on a founder's public Substack, newsletter, or long-form publication.
How we detect it
Public subscriber counts, public post engagement counts, cross-reference against founder identity verified against the company entity.
Why it matters
Founders with growing public audiences have asymmetric distribution. They can launch and scale things with less paid acquisition than competitors. The audience is itself an asset.
Founder Twitter Engagement Quality
Medium weightWhat it measures
Engagement quality on a founder's public social posts — not raw follower count, but the depth of interaction from recognised practitioners in the founder's category.
How we detect it
Engagement graph analysis on public posts: who replies, what their profile says they do, whether they're cited elsewhere in the category.
Why it matters
A founder whose threads pull in genuine domain practitioners has a different category position than a founder whose threads pull in only followers. The first is a category authority; the second is an audience.
02Scoring formula in plain language
A company's total Foxhole score is a weighted sum of its active signals, with recency decay, and a multiplier that applies only under specific combinations.
03Capital Efficiency Score
The Capital Efficiency Score (CES) measures how much real-world output a company appears to produce per unit of headcount. It runs alongside signal scoring and contributes a small adjustment to the final number.
Inputs
Public headcount estimates from major directories, the company's own pricing surfaces (where published), and public traffic indicators. No private revenue data — ever.
Confidence bands
Every CES score carries a confidence band reflecting how complete the input picture is. A company with public pricing, public team page, and public traffic data scores at higher confidence than a company with only one of those.
Tiered output, not point estimates
We surface tiers (Exceptional / Strong / Average / Below average / Insufficient data) rather than a precise revenue-per-employee number, because the uncertainty in any single component makes a point estimate misleading.
| Tier | What it means |
|---|---|
| Exceptional | Public surfaces suggest the company produces more output per employee than the strong majority of comparable companies in its sector. |
| Strong | Above the median of comparable companies; the company appears to be using each headcount efficiently. |
| Average | In line with comparable companies in its sector and stage. |
| Below average | Public surfaces suggest output per headcount lags comparable companies. Could be true; could also mean public surfaces are out of date. |
| Insufficient data | Not enough public surface area to produce a confident score. The CES modifier does not apply at this tier. |
Why tiers and not exact revenue figures: any inferred revenue number is a guess wrapped in math. Publishing it as a precise figure invites legal risk and gives investors false certainty. Surfacing a tier with a confidence band tells the truth: "here is what the public surface suggests, and here is how confident we are in that suggestion."
04What we don't do
The negative space matters as much as the positive. These are categories of intelligence that other tools build and Foxhole deliberately does not.
- We don't track founder sentiment or psychology. No tone analysis on personal posts.
- We don't predict liquidity intent from personal social activity. Whether a founder is "thinking about an exit" is not a signal we infer from posts.
- We don't sell your watchlist or thesis data. Your private theses stay yours.
- We don't surface companies behind paywalls or NDAs. Everything we score is publicly verifiable.
- We don't scrape private databases, leaked CRMs, or hacked datasets. Public sources only.
- We don't ingest data from employee tracking, browser fingerprinting, or behavioural surveillance.
05Data sources
Every signal Foxhole produces is traceable to a publicly verifiable source. The categories we draw from:
- Public GitHub repository activity — commits, contributors, stars, organisation membership.
- Public job postings on major boards and company career pages.
- Public press releases, product launches, and corporate communications.
- Public corporate registries, including international filings for entities outside the US.
- Public social posts from verified founder and company accounts.
- Public product community surfaces — ProductHunt, Hacker News, technical forums.
- Public technical blog posts, conference talks, and engineering deep-dives.
05bReceptiveness Radar
Foxhole's signal score answers should I reach out. Receptiveness answers a different question — when. Five operational, observable signal sources, each completely optional. Receptiveness is never folded into the main signal score; it is a separate chip and a separate sort option.
What feeds the Receptiveness Radar
- Posting velocity change. Public post count on the founder's verified social account over the last 30 days versus the prior 60-day baseline. A positive delta is a signal. A quiet period is never used to penalise — quiet is normal.
- Executive hire recency. The most recent senior hire on the company (executive function, enterprise, compliance) drawn from the same public job-posting and announcement feeds that drive the main signal layer. A fresh hire is a bandwidth signal — the company has just expanded its capacity to take new conversations.
- Product release cadence change. Shift in public commit and release cadence on the company's public engineering surfaces. Acceleration is a signal of momentum; deceleration is never used to penalise.
- Advisory role added recently. A new advisory or board affiliation observable from the founder's public profile in the last sixty days. A new advisory is a public commitment of attention to a new surface — a bandwidth and openness signal.
- Conference and podcast appearance count. Public speaking engagements, podcast guest appearances, and conference talks in the last ninety days, deduplicated by domain. Founders who are publicly active are easier to reach.
What we don't do
The Receptiveness Radar is deliberately narrow. The following are out of scope — permanently.
- No personality inference. We never characterise a founder as introverted, extroverted, agreeable, disagreeable, or any other trait label.
- No mood scoring. We never claim a founder is tired, burnt out, frustrated, excited, energised, or in any other emotional state.
- No psychological labels. We never apply any clinical, sub-clinical, or pop-psychology category to a founder.
- No intent inference. We never claim a founder is exiting, fundraising, hiring secretly, or planning any other action that has not been publicly announced. We observe events. We do not predict the events behind the events.
- No liquidity-decision prediction. Receptiveness is not a signal that a founder will sell, raise, or accept a strategic check. It is a signal that the door is more likely to be answered. Different question.
- No sentiment analysis. We never read the content of a founder's posts to assign positive, negative, or neutral sentiment. Volume and frequency only.
How we interpret tiers
High receptiveness appears when multiple operational signals indicate an open window — for example, a fresh senior hire combined with elevated posting activity and recent public appearances. The recommended frame is "reach out this week, not next month." A high tier is a description of the founder's current observable activity pattern, not a prediction that they will engage with any specific message.
Receptive (medium tier) appears when some operational signals are present but the picture is mixed. The recommended frame is "reasonable to reach out within the next two weeks." Combine with the company's main signal score and your own thesis fit.
Low and insufficient-data tiers are never shown publicly as chips. A low tier means the founder does not currently show the observable activity pattern that correlates with open windows — it is not a judgement of the founder or the company. Insufficient-data means we could not find enough public signal sources to score with reasonable confidence.
Confidence scales with sources. When all five operational signal sources return data, confidence is high. When only one or two return data, confidence is low and the score is suppressed to insufficient-data regardless of the underlying values. Receptiveness scores expire after fourteen days and refresh on the daily scoring cycle.
06Confidence and limitations
Honest framing of what Foxhole is and is not.
We're new. Foxhole scoring improves with longitudinal data. Companies that have been on the radar longer have richer signal histories and more reliable scores. The first three months for any single company are the noisiest.
We surface the methodology so you can challenge it. If you think a weight is wrong, tell us. If you think a signal type is missing, tell us. We update the methodology in public.
If we score a company wrong, tell us. Every score is reviewable. Every signal carries its source URL. We don't ask you to trust us — we ask you to check us.
No predictions, no guarantees. A high Foxhole score is a description of a company's current public-signal pattern, not a prediction of outcome. Companies with high scores can fail. Companies with low scores can win. The score is a sourcing prior, not a forecast.
See it on a real company
Open the sample report to see how this methodology surfaces as a four-section sourcing brief, or request access if you want the live feed for your thesis.