The five classes of leading indicator

Every forward signal a company gives off — a hiring surge, a plant under construction, a drug filed with a regulator, an order won, a spike in the e-way bills it generates — is trying to tell you the same thing: revenue is coming that the financial statements have not yet recorded. Module 087 established that the single best signal to watch is a different thing in every sector. This module organises that scattered map into a framework: every one of those signals belongs to one of just five classes, and once you can name the class, you can reason about any of them the same way.

The five classes are people, physical capacity, R&D and pipeline, contracts and order flow, and external and primary data. For each, three questions have the same shape in every sector it applies to: what does it lead, how much lead time does it give, and what is its main failure mode — the way a signal that fired can still come to nothing. Getting those three right, class by class, is the equipment for the rest of Part Eight.

A list of indicators — order book for capital goods, dealer adds for autos, filings for pharma — is useful but brittle. It tells you what to watch in the sectors you have already memorised and abandons you the moment you pick up a business that fits none of them. A framework of classes does not: confronted with an unfamiliar company, you ask which of the five classes its growth is written in, and the reasoning transfers wholesale from a sector you do know.

The framework also carries the honesty this part is built on. A leading indicator leads; it does not promise. Between every signal and the revenue it points to sits a step where the signal can fail — and the framework makes that step explicit for each class, so the reader never mistakes a fired signal for booked revenue. That step has a name, and it is the second half of this module.

Each class sits upstream of revenue at its own characteristic distance — its lead time — and each converts into revenue only across a gate it can fail to pass.

• People. Hiring, net headcount, fresher intake, the provenance of senior lateral hires. It leads revenue in businesses where people are the capacity — services, IT, staffing. Lead time is short, one to three quarters. Found in headcount disclosures and, live, on public job portals. Its failure mode: hires that sit on the bench, added faster than they are deployed — headcount up while utilisation falls, so the cost lands and the revenue does not.
• Physical capacity. Capex, capital work in progress, and capacity measured in units, tonnes or beds. It leads output in asset-heavy businesses — cement, metals, hospitals, hotels. Lead time is long, four to twelve quarters across the build and the ramp. Found in the cash flow statement, the CWIP line, and capacity disclosures. Its failure mode: capacity built ahead of demand — the commissioned plant stuck at low utilisation, the J-curve that never turns.
• R&D and pipeline. Product filings, ANDAs, patents, approvals, the launch calendar. It leads new-product revenue in research-led businesses — pharma above all. Lead time is the longest of the five, two to four years. Found in regulator databases outside the accounts entirely. Its failure mode: the filing rejected or delayed, or the launch met by price erosion so fierce the approved product earns almost nothing.
• Contracts and order flow. The order book, pre-sales, and regulatory approvals that book future revenue. It leads revenue by the execution period in businesses that sell contracted work — capital goods, EPC, real estate. Lead time is two to nine quarters. Found in exchange order-win filings and backlog disclosures. Its failure mode: the order cancelled or re-priced, or won at a thin or loss-making margin — a backlog that converts to revenue but not to profit.
• External and primary data. Public hiring portals, satellite and traffic imagery, GST and e-way-bill volumes, vehicle registrations, app downloads, and old-fashioned channel checks. It leads demand broadly and in near-real time, across almost any sector — most powerfully consumer and autos. Lead time is short to nil, zero to two quarters. Found entirely outside the company's reporting. Its failure mode: the proxy is noisy or industry-wide, so it moves without the specific company moving with it.

The recurring word is the point. Each class has a conversion gate — the deploy, the fill, the approval, the execution, the map-to-this-company — and the discipline is to attach a rough conversion rate to every signal rather than treating it as revenue already in the bank. A signal that leads by a year still earns nothing if it never passes its gate.

Take one company and sort its signals into the five classes. Meridian Systems illustrative, a composite mid-cap IT services firm, reports a soft-looking quarter — revenue up just 4%. [illustrative] Rather than read the lagging top line, sort the forward signals by class.

People: net headcount rose 9% and fresher intake doubled, with several senior laterals hired from a larger rival — the dominant class for a services firm, and it is pointing up. Contracts: total contract value signed in the quarter grew 22%, a second forward class confirming the first. External data: the firm's job-portal postings are running at a record, an outside read that corroborates the internal headcount number. Three of the five classes are lit and agreeing; the 4% revenue is the rear-view mirror.

But apply the gates before celebrating. The people signal converts only if those hires are deployed — so check utilisation, which has slipped from 84% to 80%, meaning some of the 9% is sitting on the bench, not yet billing. The contracts signal converts only over the execution period and only at a decent margin — so read whether the 22% TCV is net-new or padded with renewals, and at what price it was won. The signals lead; the gates decide how much of the lead becomes profit. Meridian looks like a firm growing faster than its printed revenue — but the confidence in that read comes from naming the classes, checking that several agree, and pricing each one's conversion rather than trusting the surge.

Here is the inversion, and it is the reason the framework exists rather than a single rule. All five classes exist for almost every company, but which class is the dominant, most-reliable lead differs by sector — and pointing at the wrong class is watching a signal that was never going to move the business. People lead an IT firm; physical capacity leads a cement maker; the pipeline leads a pharma company; contracts lead capital goods; external data leads a consumer or auto business whose demand is too diffuse to read from any single internal line.

The five classes also line up as a reference grid — what each leads, its rough lead, where to find it, and the gate at which it can fail.

Sorting a signal into its class tells you what to watch and how much lead it gives; it does not tell you the signal will convert. The conversion rate is the framework's honest limit — an order book is future revenue times a win-and-execute rate, a filing is a launch times an approval rate times a price, a hiring surge is revenue times a deployment rate. The class tells you the numerator; only judgement, history and the gate give you the fraction that reaches revenue.

It cannot give you a precise lead time. The ranges here are indicative and stretch in practice — an approval cycle slips, a plant's ramp drags, an execution period extends when a client defers. And it cannot rank a company's classes for you: knowing that people dominate IT and capacity dominates cement is the pattern, but a specific company can be the exception — a services firm buying growth through acquisition rather than hiring, a cement maker whose next leg is a competitor's distress rather than its own capex. The framework points your attention at the right class; it does not do the judging inside it.

The first trap is treating a fired signal as booked revenue — reading a commissioned plant, a fat order book or a doubled fresher intake as growth already in hand, skipping the gate. The signal is the numerator; the conversion rate is what turns it into revenue, and ignoring the gate is how investors pay today for growth that never arrives.

The second is watching the wrong class for the sector — importing the contracts class into a cement maker that has no order book, or the people class into a capacity business where output is set by kilns, not headcount. A signal read in the wrong class produces confident nonsense: it moves, you react, and it was never the lead for that business.

The third is trusting one class alone. A single class lit is a question, not an answer; the confident read is several classes agreeing — hiring and deal wins and portal postings all pointing the same way for an IT firm. And within a class, a single data point is rarely the signal: one month's job postings or one quarter's order inflow is mostly noise; the trend across several is the signal, and reading a single lumpy number as a turn is how a reader gets whipsawed by a series that was always going to bounce.

The fourth is forgetting the second gate on the longest-lead class. In pharma the pipeline must pass approval and survive price erosion; an approved product in a collapsing-price segment earns a fraction of what the filing seemed to promise. A class with two gates converts at the product of two fractions, and pricing only the first overstates the lead badly.