What a hiring signal actually tells you

One post, a dozen data points

Most descriptions leak a couple of useful facts. Occasionally one leaks everything. The clearest example I found was an SDR posting from an HR-tech company of around 130 people, and it opens like this.

One sentence, and you already know a US expansion is underway and the outbound function starts at zero. The rest of the post keeps giving.

Now read it as three different sellers. If you sell GTM tooling, this is a company mid-build with a named stack to displace or complement. If you sell into HR or procurement, they just told you their ICP is your ICP. If you compete with them, you know they're profitable, growing 60% a year, and entering the US with exactly one SDR. A competitive briefing, published as a job ad.

That's the thesis of this whole piece in one post. A job description advertises a role for a few weeks. It broadcasts the company's operating state to anyone who reads past the title.

The gems catalog

Ten things worth extracting, each with a real excerpt from the sample and the read that follows from it.

The honest coverage picture

Gems are only worth building on if they're there when you need them, so I measured how often each dimension actually shows up across the 411-post base.

The practical read: the top six dimensions are dense enough to build on. The middle ones work as enrichment, present often enough to add real value and sparse enough to need a when-available disclosure. The bottom ones, named leaders, customer name-drops, compliance posture, are gold when they land and shouldn't anchor any core surface.

A few of the middle dimensions deserve a sentence each. Compensation structure fingerprints org type: startups write equity and profit share, growth SaaS writes OTE and RSUs, PE-owned writes base plus bonus. Benefits are a procurement inventory, because every named benefit is usually an HR-tech vendor relationship. And the GTM role mix over 90 days beats the funding stage as a stage indicator: three SDRs and no CSMs is a company building pipeline, five CSMs against one AE is a company defending its base.

Hard numbers: the sparse-but-gold tier

After the first pass I went back and grepped the sample for hard dollar figures specifically: revenue, quota, ARR, targets. They exist, and one post was the richest sales-KPI record in the entire corpus.

That's an AI company of around 630 people publishing individual-rep productivity, quarterly pace and a third-party sales-org ranking in a job ad. If you sell sales tooling, that's board-level intel. If you're a candidate benchmarking compensation, it's the number the recruiter would never volunteer.

The rest of the tier looks like this. A payments platform of around 1,400 people: "we serve 3,000+ customers worldwide, managing $15 trillion in payments annually." A three-person beverage startup: "launched in 315+ stores of a national retailer in March, with sales growing 40% month over month." A health-IT company of around 220 people: "over 2,300 customers" with an OTE of $200k+ printed next to it. And the healthcare-data post from gem two, with its $10M P&L and 25% growth target stated outright.

The honest coverage read on all of it:

• Revenue and business-size dollar figures: roughly 4 of 411 posts, about 1%.
• Quota or OTE numbers: 2 posts.
• Customer or user counts: around 7 posts, about 2%.
• Growth percentage targets: around 4 posts.

So you can't build a filter on this tier, it's far too rare. But when a post carries a revenue number, it's often the most valuable single fact in the whole record. The pattern behind where they show up: recruiting-brag posts written to sell the company to candidates, and senior revenue-leadership descriptions where the P&L scope is stated to attract the right seniority. Both are written by people close to the numbers, which is exactly why the numbers leak.

The second pass: what I missed first time

I went back through the same window with a looser threshold, 434 posts at over 300 characters, looking for dimensions the first pass skipped. Several are stronger than anything above, and two of them no other B2B data source surfaces at all.

Advisory-board build-outs. One record showed a company with two employees that had posted fifteen roles in the window. That reads as an absurd velocity outlier until you look at the titles: endocrinology advisor, obstetrician-gynecologist advisory board member for women's metabolic health, hepatologist covering MASLD and MASH, gastroenterologist, family medicine, psychiatrist, cardiology advisor. That is not hiring. It is a two-person company assembling a clinical advisory board, and the specialty mix tells you exactly what they are building: a metabolic-disease care platform, almost certainly GLP-1 adjacent.

You can reverse-engineer an unreleased product's clinical scope from the board it recruits. And advisory-board assembly is pre-launch or pre-raise activity, because companies buy credibility before they need it. That makes it the earliest-stage signal in the whole dataset, firing before the product exists, before the raise, before any press. The generalisable rule: when roles posted over employee count blows past 100%, stop reading it as growth and read the titles. Advisory board, board of directors, fractional exec and scientific advisor postings are a distinct class, and they mark a company buying credibility rather than capacity.

The post language beats the country field. One record carries a US company country and around 340 employees. The post is written entirely in Korean, recruiting full-stack and mobile engineers in South Korea. Every firmographic database in the world will tell you that company is American. The hiring signal tells you where they are actually building a team right now. The country field describes the headquarters, the post language describes the frontier, and the mismatch between them is invisible to every static source. It fired on under 1% of posts, but script detection is trivial and the false-positive rate is essentially zero.

Shelf-life is a hiring-difficulty index. 149 postings carried both a posted date and an expiry. Most cluster at 30 or 38 days, which is just platform default and ATS config. The interesting cohort is the seven or eight running 180 days or more, and it gets sharper crossed with applicant count. One aerospace and defense company of around 470 people was running two 180-day senior engineering requisitions with under 25 applicants each. Long window plus low applicant count means a hiring function under strain, which is a qualified lead for anyone selling sourcing or staffing, and a flag for anyone modelling that company's execution risk.

Band width reads levelling maturity. 102 posts carried a parseable salary range. A global consultancy posted bands 180% and 193% wide. A fintech, a learning platform and an engineering services firm posted bands of 7 to 9%. A tight band means the company knows exactly what level it is hiring and has a real compensation structure behind it, which usually travels with formal procurement, longer cycles and more stakeholders. A very wide band means the requisition spans levels or geographies, so the posting is a funnel rather than a role.

Clearance and ITAR language, 10%. Higher coverage than anything else in the sparse tier. An export-control or clearance requirement means the company holds or is chasing US government and defense work, which reshapes everything downstream: procurement measured in quarters, FedRAMP expectations on any SaaS they buy, US-person staffing constraints. If you sell software, a clearance line tells you whether you need a government-cloud story before the first call.

AI-in-hiring disclosures, 3% and climbing. This dimension barely existed a year ago. Companies now write that they use AI tools to review applications and that those tools assist rather than replace human judgment. Two things fall out. The company is running AI in production on a compliance-sensitive workflow, so they are past experimentation. And they have counsel tracking AI-hiring regulation, so their compliance function is current. For anyone selling AI governance or model monitoring, that is a materially warmer start than cold market education. Worth tracking as a time series.

Employment structure is a runway proxy, 6%. Equity-only or commission-only compensation means the company cannot fund the seat in cash. One 30-person AI company offered a part-time advisory role at two to three hours a month for equity participation. Track the full-time versus contract ratio across an account's postings over time and you have a cash-position read that updates faster than any funding announcement, in both directions: a company converting contract roles to permanent is a company whose budget just opened.

Per-state salary tables, 6%. US pay-transparency law forced these into the open, and a handful of posts publish the full grid. One global consultancy listed thirteen jurisdictions in a single description, each with its own band, the highest floors clustered in three states and the cheapest market clearly visible. That's a map of where a company actually employs people plus the exact cost-of-labour differential it applies between markets, free to anyone doing territory planning or comp benchmarking.

Travel percentage, 5%. A stated travel figure on a GTM role tells you the motion: under 10% is inside sales, 25% or more is field sales with on-site enterprise deals. The same figure on a product or engineering role means customer-embedded development, where engineers go to sites. Different culture, different buying profile.

On-call and 24/7 language, 2%. Round-the-clock staffing is expensive, and a company only funds it when customers are contractually entitled to it. A rotating-shift mention therefore tells you the company carries uptime commitments, which makes it a candidate for observability, incident management and on-call scheduling tooling, and tells you its customers run something mission-critical on it.

Certifications and fraud warnings. Two small ones worth knowing. A required certification, 3%, proxies the regulatory regime the company operates inside: an accredited-lab standard means accredited labs, an active state accountancy licence means real statutory reporting, and both lengthen every procurement cycle. And 2% of posts carry a recruiting-fraud warning, telling candidates that offers only come from the company's own domain. Scammers impersonate employers people want to work for, so the boilerplate is indirect evidence of employer brand strength, which is an orthogonal read you won't get from headcount or funding.

The structural signals, no text extraction required

Worth separating these out, because they need no language model at all. They fall out of fields the API already returns, which makes them deterministic: no false positives, no coverage gaps beyond the underlying field. If you are productising this, ship these first and add the text mining afterwards.

Every dimension, both passes

The complete picture across both passes, sorted by how often each dimension actually appears. The tiers run bronze to gold, and they run that way because rarity and value move together here. Bronze is dense enough to build a product on, and every competitor has it too. Silver adds real value with a when-available caveat. Gold is too rare to filter on and often the most valuable single fact in the record, which is exactly why almost nobody extracts it.

The hire is a sequence, and you see one step

One caution before the stacks, because it's the mistake most hiring-intent tools bake in. A hire doesn't happen out of nowhere. What we observe in the data is a sequence of events that leads to it: a funding round closes, an executive arrives a quarter later, and the team gets hired after that. An acquisition triggers departures, and the backfills follow within two quarters.

A hiring signal is the most legible step in that chain, and it's still one step. Read alone, it tells you a seat opened. Read next to the funding round that paid for it, the job change that preceded it and the M&A that may have caused it, it tells you where the company is in its own story, which is the context the message actually needs. That's why everything below assumes the signals sit stacked on one company profile rather than in separate tools.

Where hiring signals get powerful

A single post is a data point. The unlock is stacking it against the other signals on the same account.

Silence is the stack nobody runs. For a funded company, days since the announcement without a hiring signal is a distress indicator. Our funding-to-staffing study put the median gap at 35 days. A Series B sitting at day 90 with zero hires isn't neutral. Plans changed, the round is being renegotiated, or the go-to-market thesis fell apart internally. That read only exists when funding and hiring sit in the same view.

Repost rate is the other one. In a dedup pass over 242 feed posts, 55% were reposts of identical text. Heavy repost density on one company means they're leaning on external recruiters, which is a product-fit signal for talent tech and a market-share signal for the agencies. You can count who's reposting and rank them by activity.

Reading job descriptions at scale

The company-authored intro beats the requirements list. Everyone reads the bullet points. The value sits in the about-us and why-this-role paragraphs, where companies editorialise: growth rates, customer counts, admitted gaps, analyst wins. The requirements describe the candidate they want. The intro describes the state of the business.

Enterprise posts and startup posts leak different things. Big-company postings leak process: application windows, compliance language, structured compensation, the manual work they want automated. Startup postings leak strategy: growth rates, first-hire status, the exact stack, named executives. Read them differently.

The brag and the tell travel together. We're growing 60% sits in the same post as we don't have a playbook. The brag qualifies the account. The tell shows you where they hurt. Read both.

What to build off this

If you're operationalising hiring signals, these are the extractions worth prioritising.

What only hiring signals give you

Funding signals tell you a company has money. Job-change signals tell you someone moved. Hiring signals tell you what capability the company has decided to build next.

That last one is the sharpest predictor of what a company is about to spend on, because it's a public commitment rather than an inference. The seat was agreed internally, the headcount approved, the budget cleared and the description published. Everything they need to buy to make that hire successful is now an open window measured in weeks.

You can't get that from a press release. You can't get it from technographics. You can barely get it from intent data. It's sitting in the description text of the posts we argued were arriving too late, and almost nobody reads it deeply. What to write once you have the detail is in the copy piece.