CWIP and the understated denominator

A company is building something big. On the balance sheet it shows up as capital work in progress - money already spent on a plant, a line, a project that is not yet finished and not yet earning. You are looking at two things at once, and they pull in opposite directions.

Seen one way, that CWIP is the future arriving early: capacity you can watch being built quarters before it commissions, growth made visible on the balance sheet before the P&L has any idea it is coming. Seen the other way, it is a lump of capital that earns nothing yet but still sits in the denominator of every return ratio you compute - so a company in the middle of a genuine, value-creating build reports a return on capital employed that looks worse than the assets actually running deserve. And a third possibility sits underneath both: management can leave costs in CWIP on purpose, precisely because CWIP does not depreciate, using construction as a place to hide expenses that would otherwise hit profit.

This module is about reading all three at once - the leading indicator, the distortion, and the forensic tell - from the same line.

The mechanics that create the ambiguity are simple, and they follow directly from the capex cycle (086) and the J-curve (098). When a company spends on a long-lived asset, the spend does not go through the P&L; it is capitalised onto the balance sheet. While the asset is still being built, it sits in CWIP. Only when it is finished and ready for use is it transferred out of CWIP into property, plant and equipment - at which point it starts to earn revenue and, for the first time, starts to attract depreciation.

Three consequences flow from that single rule, and each is one of the three readings:

• It leads. CWIP is spend committed to capacity that has not yet arrived. A rising balance is future output being built, visible before any of it shows in revenue - the earliest place on the statements to see growth coming.
• It drags. CWIP is capital employed that earns nothing. Put it in the denominator of ROCE or ROE and the ratio falls, not because the running business got worse but because the base swelled with a non-earning plant. The bigger the build relative to the existing assets, the harder the drag.
• It hides. Because CWIP does not depreciate and does not hit the P&L, it is a natural place to park costs. Interest, overheads, even expenses that should have been written off can be capitalised into a project and left there, keeping them out of profit and postponing the depreciation that capitalisation would eventually trigger - a balance-sheet game of exactly the kind Part Four taught (057).

Start with the distortion, because it is arithmetic and you can undo it. Reported ROCE is operating profit divided by all the capital employed - and during a build, that capital includes the CWIP. Since the CWIP contributes nothing to the numerator, the honest way to read the running business is to strip it out and compute the return on operating capital: the same EBIT, divided by the capital that is actually operating.

That recomputation is the first move, and it is the one the glossary already points you to - CWIP in the denominator understates a mid-build company's returns, so recompute excluding it (071). But stripping CWIP tells you what the running assets earn; it does not tell you whether the CWIP itself is real. For that you need two more reads:

• The CWIP ageing. How long has the balance sat before being capitalised? A build has a known gestation - a cement line takes two to three years, a power plant longer. CWIP that moves toward commissioning on roughly that schedule is a genuine project. CWIP that sits, barely changing, across three or four annual reports without ever transferring to fixed assets is the forensic tell: a stalled project, or costs parked to defer depreciation. The notes to the accounts increasingly disclose a CWIP ageing schedule and projects overdue against their original completion date - read it.
• The CWIP-to-gross-block ratio. CWIP measured against the original cost of the assets already in place tells you how large the build is relative to the running business. A ratio near the size of the existing gross block says the company is roughly doubling its capital base - real ambition, and a large future depreciation and interest bill about to land. The ratio's level is not itself good or bad; it is the context for the ageing read.

The level of the CWIP balance is close to noise - a big number is what any real expansion looks like mid-build; the signal is whether it is moving toward commissioning, which only the ageing and the year-on-year transfers to fixed assets reveal.

Here the same line inverts. A large standing CWIP balance is normal, expected and healthy in a long-gestation builder, and anomalous, even alarming, in a business that has no reason to be building for years. What counts as patience in one sector is negligence in another.

The discipline is never to read a CWIP number in isolation. Ask what the sector's normal gestation is, and judge the balance and its ageing against that - the identical figure that earns a cement maker your patience should cost an FMCG company your suspicion.

Take two composite companies that show the same distorted number for opposite reasons.

The first, a cement maker we will call Deccan Cement illustrative, reports ROCE of 11%, down from 17% two years ago, while its peers earn 17–18%. [illustrative] An investor screening on the raw figure rejects it as a business that has gone ex-growth. Open the balance sheet: CWIP has risen to ₹400 cr against ₹600 cr of running assets - a new line, two-thirds the size of the existing plant, under construction. Strip it out and the same operating profit is earning about 18% on the assets actually running, comparable with the peers who simply are not mid-build. [illustrative] Check the ageing: the CWIP has grown each year and the notes show the line on schedule for commissioning next year. This is the distortion, not decay - the return will step back up when the plant transfers to fixed assets and starts earning on the full base. The depressed ROCE was the price of the build being visible before the output.

The second, an asset-light packaged-foods company we will call Saffron Foods illustrative, also carries a CWIP balance large against its modest fixed assets - and its reported ROCE is likewise dragged down. [illustrative] But strip the CWIP and recompute, and you have answered only half the question. The forensic half: this is a business that should not need a large multi-year build at all, and the balance has sat, barely moving, across three annual reports without being capitalised. [illustrative] The ageing schedule shows a project long overdue against its original completion date. Here the same understated denominator is not patience rewarded but a warning - a stalled diversification, or interest and overheads being capitalised into a project that never commissions, keeping them out of the P&L and deferring the depreciation that transfer would bring. Same line, same arithmetic drag, opposite verdict.

The recomputation has real limits. Stripping CWIP tells you the return on the assets running today; it says nothing about whether the plant under construction will earn that same return once it commissions. A company can be earning 20% on its existing assets and building a new line that will earn 8% - the return on operating capital flatters the incremental project. The last cycle's return is the best forecast of the next (086), but it is a forecast, not a fact.

CWIP also cannot, by itself, tell you why a balance is ageing. A project can stall for honest reasons - a delayed environmental clearance, a contractor dispute, a funding crunch - that have nothing to do with hiding costs. Ageing CWIP is a question, not a conviction; it tells you to read the notes, the management commentary and the interest-capitalisation policy, not to conclude fraud.

And the line does not reveal what has been capitalised into it. Whether interest and overheads have been swept into the project, and on what basis, sits in the accounting-policy notes, not in the CWIP figure. The number tells you a build is underway; the notes tell you what the build is made of.

The first trap is taking the reported ROCE at face value during a build - screening out a genuine compounder because the mid-build denominator drags its return below a threshold, or ranking a mid-build company below a peer that simply is not investing. The raw ratio is not comparable until the CWIP is stripped; judging the two side by side without adjusting compares a company mid-investment against one standing still.

The second is the mirror image: treating every large CWIP as a genuine build. Because the honest distortion and the dishonest hiding place look identical in a single year's figure, an investor who has learned to strip CWIP and admire the recomputed return can wave through a parked-cost balance without ever checking whether it commissions. The recomputation answers the distortion; only the ageing answers the forensic question.

The third is reading the CWIP level as the signal. A big number triggers alarm and a small one comfort, when the level is mostly context - a large balance is what any real expansion looks like. Whether the balance is moving toward commissioning, on the schedule the sector implies, is the actual signal, and it lives in the ageing and the year-on-year transfers to fixed assets, not the headline figure.

The fourth is forgetting the bill that commissioning brings. Investors cheer a plant switching on and forget that the day CWIP transfers to fixed assets is the day depreciation begins and, if the build was debt-funded, interest stops being capitalised and starts hitting the P&L. A genuine build lifts the return eventually, but the first effect of commissioning can be a step-down in reported profit as those charges land - the J-curve (098) the previous module taught, now visible in advance on the CWIP line.