Capacity in physical units, not rupees

A company announces a ₹1,800 crore expansion. The headline travels — the number is large, the intent is growth, the market nods. But that rupee figure cannot be turned into a single rupee of forecast revenue, because it answers the wrong question. It tells you what the expansion cost; it does not tell you what the expansion can produce. And revenue is produced, not spent.

This module is about the one input that does forecast revenue: forward capacity read in physical units — tonnes for a cement plant, megawatts for a power station, rooms for a hotel, spindles for a spinner, litres a day for a dairy, subscribers for a telco. Capacity in units, multiplied by a realistic utilisation and a sober realisation, gives you a bottom-up revenue forecast years ahead — one you build yourself, independent of what management guides. Capacity in rupees gives you nothing you can forecast with. The whole module turns on preferring the first number and distrusting the second.

A rupee of capex is a bundle of things that have nothing to do with output. It contains steel and cement prices that may have risen 30% since the last plant was built; it contains land, which produces nothing; it contains financing and pre-operative costs; it contains the difference between a company that bought its equipment well and one that overpaid. Two identical plants — same tonnes, same technology — can be built for wildly different rupee figures depending on when and how they were bought. The rupee number therefore conflates cost inflation with capacity, and there is no way, from the outside, to separate them. You cannot divide revenue by a rupee of capex and get anything meaningful, because the denominator is contaminated by everything that is not output.

The physical unit is clean. A tonne of cement capacity is a tonne whether it cost ₹4,000 crore or ₹6,000 crore to build. It maps directly onto a quantity that can be sold, and a quantity sold, priced, becomes revenue. That is why the unit is the input a forecast is built from and the rupee is not.

This is why the input belongs to a book about reading, not modelling. You are not building an elaborate model; you are choosing the right raw number to start from. Get that choice wrong — start from rupees — and no amount of modelling downstream can rescue the forecast, because the rupee figure is the noise the announcement broadcasts, and the physical capacity is the signal it often buries.

The forecast is a chain of three physical or near-physical terms, and it is deliberately simple:

Capacity × utilisation × realisation ≈ revenue.

• Capacity, in units. Find both the installed capacity and the under-construction capacity, in the sector's own unit. Both are disclosed — in the annual report's operational review, in the investor presentation's capacity table, sometimes in a plant-wise footnote. Add the under-construction capacity at its commissioning date; that is the forward number.
• Utilisation, ramped honestly. A new plant does not run flat out on day one. It ramps — perhaps 55% in year one, 70% in year two, 85% at maturity — and the ramp can stall if demand is not there. The realistic ramp, not the nameplate, is what turns capacity into volume actually sold. This is the fragile term, and the one to be conservative on.
• Realisation, sober and mid-cycle. Price per unit — revenue divided by volume. Use a realisation through the cycle, not the top of it, and remember a rising realisation is a question about its source, not a free assumption.

Multiply the three and you have a revenue forecast for a year several years out, built from numbers you sourced and assumptions you chose.

The discipline is that every term is either disclosed (capacity, realisation) or estimated by you on visible evidence (the ramp). Nowhere in the chain do you take management's revenue guidance as an input. That independence is the point: you now have a number to check guidance against, rather than a guidance figure to believe.

Take a composite mid-cap cement maker, Meridian Cement illustrative. [illustrative] Its investor presentation shows installed capacity of 2.0 million tonnes and a new line of 1.0 million tonnes under construction, commissioning in eighteen months — so forward capacity is 3.0 million tonnes. Management guides to "strong double-digit revenue growth" but attaches no number you can pin down. So you build your own.

The new line will not run flat out immediately: you assume it ramps 55% → 70% → 85% over three years, and the existing plant lifts from today's ~82% to around 85% as well. At maturity the whole 3.0 million tonnes runs near 85%, so roughly 2.55 million tonnes get sold. Realisation across the last three years has averaged near ₹5,200 a tonne, and you use that mid-cycle figure rather than the current spot, which happens to be high. The arithmetic — 2.55 million tonnes at ₹5,200 — lands at about ₹1,326 crore of revenue at normal utilisation, up from roughly ₹850 crore today. [illustrative]

Now the ₹1,800 crore capex tells you nothing you could not do without. What it does do is let you sanity-check the returns: ₹1,800 crore spent to add perhaps ₹476 crore of revenue and, at a plausible margin, some ₹90–110 crore of profit is a return you can weigh — but the revenue forecast itself came entirely from the tonnes, the ramp and the price. When Meridian's actual quarterly volumes start printing, you compare them against your ramp, not against the capex, and you learn early whether the plant is filling or stalling.

The method is fixed; the unit is not. This is the inversion at the heart of the module: the right physical unit differs entirely by sector, and reading the wrong one — or reaching for rupees because you do not know the right unit — is where the forecast breaks. Tonnes for cement and steel. Megawatts for power. Rooms times room-rate for a hotel. Spindles for a spinner, subscribers for a telco, litres a day for a dairy. And for a services or IT business there is no physical unit at all — no tonne, no megawatt, no room — so the capacity that produces revenue is headcount, and the whole physical bridge is replaced by billable people times utilisation times realisation per head. Point at the wrong unit and you will forecast the wrong plant.

Set the units side by side and the discipline becomes concrete — the same three-term chain, wearing a different first term in each business:

The bridge forecasts revenue, not profit. Capacity times utilisation times price stops at the top line; whether that revenue drops through to earnings depends on the cost structure, operating leverage and the margin the plant runs at — none of which the capacity figure contains. A plant can fill to nameplate and still lose money if it commissioned into a price war.

It also cannot tell you whether the utilisation will be there. Capacity is a supply fact; utilisation is a demand outcome, and demand is the thing the plant does not control. Where a whole sector adds capacity at once — the classic capex cycle — every player's ramp assumption fails together, and the disclosed tonnes ramp into a glut rather than into revenue. The capacity number is reliable; the utilisation assumption you multiply it by is the judgement, and it is only as good as your read of sector-wide supply and demand.

And it says nothing about timing slippage. Commissioning dates slip — environmental clearances, equipment delays, funding — so the "eighteen months" is itself an assumption. A forecast built on capacity commissioning on schedule inherits every delay the schedule hides.

The first and largest trap is reading capex in rupees as if it were capacity. A big rupee number feels like big growth, and the mind scales revenue up with the spend. But the rupee figure is mostly cost — inflation, land, financing — and in a period of high input prices a company can spend a great deal to add rather little capacity. The rupee grows; the tonnes may barely move. Anchor on the spend and you forecast growth that the physical output will never deliver.

The second is using nameplate utilisation instead of a realistic ramp. Multiplying full capacity by full utilisation on the commissioning date produces a forecast that assumes a new plant is sold out from birth. It never is. This is where rejected guidance sneaks back in — you disowned management's revenue number but adopted their optimism as a 100% utilisation assumption, and arrived at the same inflated place by a different road.

The third is reaching for the wrong unit, or for rupees when you do not know the unit. Someone forecasts a power company on megawatts of nameplate and forgets the load factor; someone forecasts a hotel on rooms and forgets occupancy ramps for years; someone, faced with an IT firm that has no physical unit at all, gives up and scales revenue off the rupee capex — the one number that forecasts nothing. The unit inverts by sector, and using the wrong one is using the wrong plant.

The fourth is treating your own bottom-up number as more certain than it is. The forecast is honest arithmetic on top of one fragile assumption — utilisation — and dressing it up as precision ("₹1,326 crore in FY-three") hides that fragility. It is a sober central case built independently of guidance, and its value is exactly that independence, not a false decimal-point accuracy.