Public hiring data as primary research

A company tells you what it did once a quarter, in a filing that closed weeks before you read it. But every day, in public, it tells you something about what it intends to do next: it advertises the jobs it is trying to fill. The count of open roles, and — more revealingly — their type, where the functions, skills and locations it is hiring for, sit on public job portals and on the company's own careers page, visible to anyone, months before any of it reaches a financial statement. The previous module taught you to read headcount as it appears in the accounts — a reported number, backward-looking, easy to misread. This module goes outside the accounts entirely, to the hiring a company is doing right now, in the open, and asks what it forecasts. illustrative

This is scuttlebutt applied to the labour market. It carries the same logic as the dealer check or the store visit: an observable fact about the real world, gathered independently of the company's own narrative, read against a thesis you built from the numbers. A firm that is quietly staffing up three new delivery centres is building capacity ahead of revenue it can already see; a firm that has just pulled every open requisition and frozen hiring is bracing for a slowdown it has not yet announced. Neither shows in last quarter's accounts. Both are legible, in public, to a reader who knows to look — and both lead the financials by a knowable margin.

But the method carries the same two hazards as all scuttlebutt, sharpened here by how easy the data is to gather badly. The first is that a raw count of job posts is mostly noise — backfill, re-posts, seasonal bulges — and separating the real expansion signal from the churn is the entire skill. The second is a hard line: public, observable postings are fair game; data pulled from behind a login in breach of a portal's terms, or a confidential unpublished hiring plan handed to you from inside, is not — the first is a terms-of-service problem and the second is material non-public information, which is a criminal matter, not an edge. And how much the whole exercise is worth, before any of that, depends on the sector — because how tightly a company's roles map to its revenue is not the same in a software house and a cement plant.

The financial statements are, by construction, a report on the past, and headcount as it appears in them is the most backward-looking labour number of all: it counts the people already hired and already billing at the balance-sheet date. By the time a hiring wave shows up as higher employee cost, or as the revenue those people generate, the decision that caused it was taken quarters earlier — the requisition opened, the offer made, the person onboarded and ramped. The public job posting sits at the very front of that chain. It is the intention made visible before the hire, before the cost, before the revenue — the earliest observable point in the company's own labour cycle, and it lives entirely outside the accounts, on portals and careers pages the company does not control the framing of.

This matters because the alternative sources are worse. The company's own commentary about hiring — "we are investing in talent," "we are being prudent on costs" — is a self-report, framed to reassure, and it lags the actual requisitions it describes. The quarterly headcount disclosure, where it exists, is the lagging number this module's predecessor taught you to read carefully. Public postings are neither: they are not the company's chosen words, and they are not a stale count — they are the live evidence of what the company is doing with its hiring right now, gathered by you, independent of its narrative. That independence is exactly what makes them scuttlebutt rather than disclosure, and it is why they can confirm or break a growth story the accounts have not yet caught up with.

Without this stance, two failures follow, and they are the labour-market versions of the scuttlebutt failures. The first is the reader who waits for the financials — who learns a company was staffing up for a boom, or freezing before a bust, only when it prints in the employee-cost line two or three quarters late, long after the signal was free to see. The second, subtler, is the reader who does look at postings but reads the gross count as gospel — who sees "500 open roles" and concludes "massive expansion," never noticing that 450 of them are the same high-churn seats re-advertised, or a campus-season bulge that recurs every year. This module installs the forward read and, in the same breath, the discipline that keeps it from becoming a confident way to be wrong.

Public hiring data has a shape, and the shape is what turns a pile of job ads into a forward read: gather what is observable now, filter the noise out of it, and translate what survives into a signal that leads the accounts by a knowable margin.

Gather from the open, systematically. The raw material is public and free: the company's own careers page, the major job portals (Naukri, LinkedIn, and the company's listings wherever they appear), and, for many firms, the postings their staffing partners run on their behalf. What you record is not just the headline count of open roles but their composition — which functions (delivery versus support versus sales), which skills (a new technology stack, a new product line), and which locations (a new city, a new plant, a new country). Composition is where the information density is: 200 billable-engineer roles across three new-city offices is a different fact from 200 replacement roles in an existing back-office. Record it consistently, at intervals, so you are reading a trend in the same source, not a single snapshot — the change over time is the signal, a one-day count is not.

Filter backfill from net-new — this is the whole skill. A job posting is a vacancy, and a vacancy can mean two opposite things: the company is adding a seat that did not exist (expansion), or it is refilling a seat someone left (backfill and churn). Only the first is a growth signal. A firm with high attrition in a particular function will post a constant stream of openings there with no change in net headcount at all — the postings measure the leak, not the growth. The tells of net-new expansion are roles in new locations or new skill areas the company did not previously staff, a step-change in count concentrated in revenue-generating functions, and postings that persist and grow rather than cycle. The tells of backfill are roles that re-appear in the same high-churn functions, the same seats re-advertised, and a count that is high but flat. You are trying to estimate the net-new component, and treating the gross count as if it were net additions is the single most common error.

Respect seasonality and sample size. Hiring is seasonal — campus recruitment, appraisal-and-attrition season, festival-linked retail and logistics ramps — so a rise against last quarter can be a calendar artefact, and the honest comparison is year on year, against the same season. And the count is a sample: a handful of postings on one portal is not a census of a company's hiring, and small numbers move around for reasons that carry no information. most short-term movement in a number is randomness wearing the costume of information — a jump from 40 to 55 open roles on one page in one month is well inside the noise, while a sustained move from 120 to 700 across every source, concentrated in delivery and new geographies, is the kind of change that means something. Weigh the signal by its size, its persistence, and how much of the company's real hiring your sources actually capture.

Translate what survives into a lead, not a level. What passes the filter is read as one of two things: a hiring surge — net-new capacity being built ahead of the revenue it will serve, which for a people business leads billing by a quarter or two — or a hiring freeze, requisitions pulled and the careers page emptied, which leads a slowdown before it prints. Neither is a promise: a surge can be over-optimistic hiring the company later regrets, a freeze can be prudent cost discipline rather than a demand cliff. The output is a forward probability to weigh against the accounts and the rest of your thesis, with a rough lead time, never a forecast to trade on by itself.

Everything above is lawful and observable, and there are two ways to leave that safe ground — one a matter of how you gather, the other a matter of what you gather — and both matter enough to state plainly. Public hiring data is exactly that: public. A job advert on a careers page or an open portal listing is something anyone can see, and counting and categorising it is ordinary primary research, the labour-market cousin of walking a shop floor or counting stock in a dealer's yard. The method's whole legitimacy rests on the information being public and observable.

The first way you leave that ground is how. Data pulled from behind a login, harvested in bulk in breach of a portal's terms of service, or scraped from a source that itself obtained it improperly, is not fair game — even if the underlying facts are "out there," the means is a contract or computer-misuse problem, and it also tends to produce a dataset you cannot stand behind. Stay with what is openly viewable, gathered in a way the source permits.

The second way, and the graver, is what. The instant the data stops being a mosaic of public postings and becomes a specific, price-sensitive fact the company has not disclosed — next year's confidential headcount plan, an unannounced mass layoff, a secret new-facility staffing budget — handed to you by someone inside, you have left research entirely and entered material non-public information. Trading on it, or passing it to someone who does, is insider trading: a criminal offence under SEBI's regulations, not a grey area and not a clever edge.

The method is constant — read a company's public hiring as a forward signal — but how much it tells you inverts by sector, because it depends entirely on how tightly a company's roles map to its revenue. Where people are the product, a role is capacity and hiring leads growth almost directly; where output is set by an asset, a role is a cost of running it and hiring says little about how much it will produce.

The inversion is that the identical signal — public open roles up sharply — is a near-direct read on future revenue in one business and almost pure backfill in another, and the difference is not the number but the sector's production function. In a software or new-economy firm, people are the capacity, so hiring them ahead of demand is exactly how growth looks before it prints, and the postings lead the billing. In a stable commodity producer, the plant is the capacity; it produces roughly the same tonnage whether the payroll ticks up or down at the margin, so its hiring is dominated by replacing the people who leave and maintaining the asset — and reading a hiring surge there as expansion is the store-check-in-a-software-business error from the last part, transplanted to the labour market. A reader who learns "hiring surge means growth" from the IT case and carries it to a cement maker will run a diligent check against an irrelevant signal and come away confidently wrong. The instruction survives the move across sectors; the weight you place on its answer must be relearned for each business, and the first question in every case is the same — do this company's roles map to its revenue, or to an asset that would produce with or without them?

Take a composite mid-cap IT services firm — call it Cirrus Digital — that has just reported a dull quarter: revenue up 3%, headcount essentially flat, management commentary careful and non-committal about the year ahead. On the filing alone it looks like a business that has stalled, and a reader who stops at the accounts would shelve it. But you know from Part Eight that the accounts are the lagging edge, and that for a people business the forward read lives in the hiring — so you go and look, in the open. illustrative

You pull Cirrus's public careers page and its portal listings, and record the composition, not just the count. Over the last ten weeks, open roles have gone from roughly 120 to roughly 700, and the mix is telling: the overwhelming majority are billable delivery — engineers, project leads, architects — not support or admin, and a large block is tagged to two cities where Cirrus has never had an office before, alongside a new cloud-migration skill set it did not previously advertise. You check the filter honestly. Is it backfill? The roles are not in Cirrus's usual high-attrition churn functions, and the count has grown and held over the ten weeks rather than cycling — this is not the same seats re-posted. Is it seasonal? It is off-cycle for campus hiring, and the year-on-year comparison, not just the sequential one, shows a step-change. Is the sample real? The rise shows across the careers page and multiple portals and the staffing partners' listings, not a spike on one page. What survives the filter is a genuine hiring surge — net-new billable capacity, in new geographies and a new skill, built ahead of revenue Cirrus can evidently already see but has not yet booked. Against the flat filing, that is a forward signal the accounts do not carry. illustrative

Now hold the discipline, because the same exercise could have broken the thesis instead — if the 700 roles had been the same support seats re-advertised, clustered in a high-churn function, spiking on one portal in appraisal season, you would have read the identical headline count as backfill and concluded nothing, and an honest reader accepts that answer just as readily. Notice, too, what you did not do. You did not call a friend in Cirrus's HR to ask for the confidential headcount plan — that would have been the one move that turns this from research into a crime, and it would have been the less useful move as well, because a leaked internal number tells you what the company already knows, while the public, filtered surge tells you something the company has not yet said out loud. And you did not scrape the portals from behind a login in bulk; you read what was openly viewable. The habit to build is the scuttlebutt habit: send the hiring read out from a thesis in the accounts, filter its answer hard for backfill and seasonality and sample, weigh it as a forward probability rather than a forecast, and stop dead at both lines — the login and the leaked plan — every single time.

Public hiring data forecasts intent to add capacity; it does not measure the outcome. A posting is not a hire — some roles are never filled, some are evergreen ads left running, some are re-posts, and a company can advertise ambition it does not execute. Even a filled role is not revenue: hiring can be into low-margin work, into a project that is later cancelled, or ahead of demand that never arrives. The output is a forward signal about what the company is trying to do, weighed as a probability, never a count you can convert into a revenue line. Treating "700 open roles" as "revenue is about to rise proportionally" is the over-read the whole method warns against.

Nor can it protect you from the sample and the season. Your sources capture some fraction of a company's real hiring, and that fraction is unknown and uneven — a firm may hire heavily through channels you cannot see, or advertise loudly while actually filling little. the signal is in the multi-year, multi-source trend, not the single snapshot — a small move on one portal in one month is randomness, and reading it as information is how the method most often misleads. And it cannot, by itself, tell expansion from backfill; that judgement rests on composition — location, function, persistence — which you must infer, sometimes wrongly, from postings that were written to attract candidates, not to inform you.

And it cannot see what it cannot see. A company that outsources delivery, uses contractors, or acquires capacity rather than hiring it can grow with no visible posting surge at all; a company under a quiet cost programme can freeze hiring for reasons that have nothing to do with demand. The absence of a signal is not the presence of its opposite. Public hiring data is one forward instrument among the several this part assembles — powerful where roles map to revenue, weak where they do not, and always a cross-check to read against the order book, the capex, the guidance and the accounts, never a stand-alone verdict. It raises the odds of seeing a turn before the financials do; it does not deliver certainty, and a reader who mistakes a filtered hiring surge for a booked result has simply found a new number to be overconfident about.

The first way people get fooled is by reading the gross count as net growth. "500 open roles" feels like a company in full expansion, but a large share of that count can be the same high-churn seats re-advertised month after month, adding no capacity at all — the postings measure the leak, not the growth. A firm with 20% attrition in a function will post a constant, alarming stream of openings there while its net headcount barely moves. The defence is to hunt for the net-new component — roles in new locations and new skills, counts that step up and hold — and to treat a big gross number as a question, not an answer, until you have subtracted the backfill.

The second way is seasonality mistaken for a trend. Hiring has a calendar — campus season, post-appraisal churn, festival-linked retail and logistics ramps — and a rise against last quarter can be the calendar, not the company. A reader who compares sequentially, quarter on quarter, will read a seasonal bulge as a growth signal and a seasonal lull as a freeze, both spuriously. The remedy is the same one that steadies every noisy series: compare year on year against the same season, look for a change that persists beyond one calendar bump, and remember that a single quarter's move in a seasonal series is mostly noise.

The third way is sector-blindness — carrying the IT-firm reading, where hiring is a clean forward signal, into a business where it is not. The reader who sees a stable commodity producer's postings rise and concludes "expansion" has run a real check against an irrelevant signal, because that plant's output is set by its asset, not its payroll, and its hiring is mostly replacement. The discipline is to ask first whether the company's roles map to its revenue at all, and to place little weight on hiring data where they do not — the same signal is worth a great deal in one sector and almost nothing in another, and confusing the two is the labour-market version of importing one sector's channel check everywhere. And underneath all three sits the oldest hazard, confirmation bias: a reader who wants the growth story will find the reassuring postings and dismiss the awkward ones, so the honest test is whether your hiring read has ever once broken a thesis you hoped was true.