How to Normalise Earnings and Cash Flow for Valuation

Normalising a year means cleaning it until what is left is the part that will repeat and the part that belongs to the business being valued. Five checks do it: separate operating from non-operating, remove what will not recur, tax the right base, establish whose profit it is, then match the capital base to the earnings measure. Sankalp Industrial Systems Limited, invented, is worked through all five.

The opening settings hold Sankalp Industrial Systems Limited's Year 0, and every figure worked below comes from them. Revenue of Rs 12,00,00,00,000 carries EBITDA of Rs 2,88,00,00,000, and nothing one-off sits inside it, so normalised EBITDA is the same figure. Depreciation and amortisation of Rs 48,00,00,000 leave operating profit of Rs 2,40,00,00,000, and tax at the company's own assumed 25.0 per cent takes Rs 60,00,00,000, giving operating profit after tax of Rs 1,80,00,00,000. On invested capital of Rs 12,00,00,00,000 that is a return of 15.00 per cent, economic profit of Rs 36,00,00,000 and earnings per share of Rs 6.90.

One idea sits underneath the five checks and each of them falls out of it. A completed year is not read in a valuation. It is grown. Somebody takes the last finished set of accounts, applies a growth rate to the top line, holds a margin, and produces five more years shaped like the first. The completed year stops being a report of what happened and becomes a template for what is assumed to keep happening, and a terminal value then capitalises the last of those for ever.

Growing a year rather than reading it is why the published accounts can be perfectly correct and the forecast built from them still wrong. How the statements were prepared and how the group was put together are settled elsewhere and assumed here. Normalisation is not a correction of the accounts; it is a correction of the use they are being put to. The accountant answered the question asked of them; the valuer is asking a different one.

What actually goes wrong when a year is grown rather than read?

Take the plainest version first, away from any company. A household works out in January what next year costs: it totals the twelve months just gone and adds a bit for prices. Inside those twelve months sits a wedding, paid for once. Left in the total and grown, it commits the household to a wedding every year for five years, and if the household is then valued on that basis, a wedding for ever after. Nobody decided this. The wedding arrived attached to the base year and nobody took it out.

The finance version is the same mechanism with larger amounts. Three things ride into a forecast this way: anything that happened once, such as a gain on selling something or a court case that settled; income from something the model is going to value separately anyway, counted twice as a result; and profit that belongs to somebody other than the shareholders being valued.

None of the three is visible in a single year, and all three are magnified by the act of forecasting. Read the year on its own and a one-off gain of a few crore is a footnote. Grow that year and the same gain appears five more times, and capitalising the fifth turns it into a permanent claim on value. The size of the mistake has little to do with the size of the item and almost everything to do with what happens to it afterwards.

Why are there five checks, and why does their order matter?

The checks run in a fixed order because each one hands the next one something it needs. Separating operating from non-operating comes first. Until it is settled which assets and which income are inside the business being forecast, there is no telling which items to test for repetition, which tax base is right, or what the capital base should contain. Running the capital check first matches capital to earnings that have not been cleaned yet.

Not one of the five changes a fact about the business. The five checks change what the model grows, and what it compares that growth against. A reader who runs all five and finds nothing to remove has still done the work, and the output is a written line saying so.

The table below is the procedure on its own, with no company in it. Each step says what to do, what to look at, and what a weak answer looks like. None argues the reasoning; where the reasoning is written down is named inside the step.

Check one: does this asset produce any of the profit about to be forecast?

Whether an asset produces any of the profit about to be forecast is the whole of check one. Not whether the asset is valuable, not whether it could be sold, not whether it appears on the balance sheet. Every asset a company holds appears on the balance sheet, so presence there separates nothing. The test is production.

Ask it of a shop. The shopkeeper runs a hardware business and also holds a small flat two streets away that nobody uses. The flat is worth money and produces none of the hardware profit. Forecast the hardware business and put the flat's value beside it, and everything has been counted once. Fold the flat into the shop's assets instead, and the shop looks worse at earning than it is.

Sankalp Industrial Systems Limited, invented, holds two things of that shape: a surplus land parcel carried at Rs 45,00,00,000 that the business does not use, and 26.0 per cent of Aruna Tooling Private Limited, also invented, carried at Rs 55,00,00,000. The tooling stake is an associateA stake large enough to give real influence over another company but too small to control it, so the group reports only a slice of that company's profit rather than absorbing its business. and it is equity accountedCarried as one line: a share of the other company's profit, with none of its sales and none of its expenses mixed into the group's own totals., so none of its sales or costs sits anywhere inside the group's operating lines.

The group's EBITDA for Year 0 is Rs 2,88,00,00,000, and neither asset contributed a single rupee of it, so neither can be forecast from it. Both leave the operating forecast, together worth Rs 1,00,00,00,000. Leaving the forecast is not the same as being worthless: each still has a value, and neither value is produced by growing an operating margin. How the two rejoin the answer is covered separately.

Every rupee amount in the body of this guide belongs to Sankalp Industrial Systems Limited and covers Year 0, its last completed year. The one-off pair the instrument offers is an illustration and is not in this company's record. Its 25.0 per cent effective rate is that invented company's own stated assumption, not a statutory rate, surcharge, cess or filing threshold for India.

Check two: which items are removed because they will not repeat?

Check two is where judgement enters, so the test deserves stating exactly. The test is not whether an item is unusual: plenty of unusual things repeat and plenty of ordinary things happen once. The test is repetition: will this item happen again, in something like the same way, in the years being forecast?

Six categories are where the answer is usually yes, this happened once, so each is worth looking for by name. A gain or a loss on disposalThe difference between what an asset actually sold for and the value it had been carried at on the books until then.. A litigation settlement, received or paid. A provisionAn amount set aside now against a cost the business expects but has not yet paid out. taken in the year or written back in it. A restructuring chargeWhat it costs to close, move or shrink part of a business, recognised in the year the decision is taken rather than spread out.. An insurance recovery. And, as a catch-all, any item whose size bears no relation to the revenue that produced it.

A year usually holds both directions, gains and charges, so the instrument at the top carries a field and a switch for each. Using only one of the two switches is dealt with further below.

Now the honest part, and most treatments skip it. The record for Sankalp Industrial Systems Limited holds none of those six for Year 0. There is no one-off item to remove: the company has none, and that is why both one-off fields open at zero. Running check two here produces a written line saying nothing was found, and that line is a real output. Inventing an item to demonstrate the method would put a fact into a base year that five other calculations rest on.

Check three: why is operating profit after tax taxed as though there were no debt?

Start with what operating profit after tax is trying to be. Operating profit after tax is the profit the business earns from trading, after tax, before anybody asks how it was funded. If it is going to sit above a capital base holding both borrowed and shareholder money, it cannot have had the cost of the borrowed part already taken out of it.

So the tax charged against it is the tax the company would bear with no borrowings at all. Take Sankalp Industrial Systems Limited's Rs 2,40,00,00,000 of trading profit and apply an effective tax rateWhat a year's actual charge works out at once it is set against the profit it was struck on, which rarely matches any headline figure. of 25.0 per cent, a rate the company assumes for itself rather than a statutory rate for any jurisdiction. Tax of Rs 60,00,00,000 comes off, and operating profit after tax is Rs 1,80,00,00,000.

The tax was levied on profit before tax of Rs 1,92,00,00,000, the same Rs 2,40,00,00,000 after Rs 48,00,00,000 of interest, so the company's real charge for the year is Rs 48,00,00,000. The two tax figures differ by exactly Rs 12,00,00,000, and that is precisely 25.0 per cent of the interest. A reader who cannot make that line balance has found either an arithmetic slip or an assumption they were not told about.

The Rs 12,00,00,000 difference belongs on the financing side. The difference exists only because the company borrowed, and its size depends only on how much it borrowed. That is what a tax shieldThe tax saved because interest is allowed as a deduction, so a rupee of interest costs the borrower less than a rupee. is. Keeping it out of the operating figure is practical rather than doctrinal: an operating profit that moves whenever the funding mix moves cannot be compared with anything, including the same company a year earlier. Where the benefit of the deduction is accounted for instead is covered separately.

Check four: whose profit is this, once it is followed to the bottom?

Ten neighbours put money into a shop together. Nine hold three quarters between them and one holds the rest. The till takings are the shop's till takings, all of them, and they get counted in full; what the nine can divide among themselves is not all of it. The gap between what the group's accounts total and what the shareholders being valued may claim is check four.

Sankalp Industrial Systems Limited holds 75.0 per cent of Sankalp Coatings Private Limited, invented, and that subsidiary is fully consolidatedEvery line of the smaller company is added into the group's totals at the full amount, whatever share of it the group actually holds.. Every line of it is already inside the group's revenue, EBITDA and profit at the full amount rather than at three quarters, so the profit at the bottom of the group's accounts is more than the profit the group's own shareholders have a claim on, and the difference is stated as a separate line. Walk it down for Year 0, one step to a line.

Spread that last line across a share count of 20,00,00,000 and earnings per share are Rs 6.90, not the Rs 7.20 the group figure would have given. That Rs 6,00,00,000 on the second last row is the slice of Sankalp Coatings Private Limited held outside the group, and it is the only row in the walk that turns on neither trading nor tax.

Rs 6.90 against Rs 7.20 is a difference of 4.35 per cent, on the line that appears in more comparisons than any other figure a company publishes. The minority line is small, sits low down and is easy to walk past. A line like that earns a named check rather than a glance.

Check five: does the capital base match the earnings measure sitting above it?

Every return is a fraction, and a fraction is only meaningful when its two halves were assembled on the same basis. Whatever was taken out of the earnings must also come out of the capital, and whatever was left in the earnings must stay in the capital. The rule sounds trivial and is broken constantly. The earnings are built by a valuer while the capital gets copied off a balance sheet, often by a second person who never saw the cleaning happen.

Sankalp Industrial Systems Limited's invested capital at Year 0 is Rs 12,00,00,00,000, and two lines carry all of it: net fixed assets of Rs 10,20,00,00,000 and working capital tied up in the trade of Rs 1,80,00,00,000. Two things on the balance sheet are missing on purpose. The Rs 1,00,00,00,000 of non-operating assets is out for the reason check one gave, and the Rs 1,20,00,00,000 of cash is out on the same reason: not a rupee of the operating profit came from holding it.

Put both back and the base becomes Rs 14,20,00,00,000 while the earnings above it do not move. Those assets never contributed to them. Measured return drops to 12.68 per cent from the 15.00 per cent it should read, wide enough to change what anybody concludes about the company, and produced entirely by an assembly error.

There is a quiet reward for anyone who runs the check properly. Revenue of Rs 12,00,00,00,000 and invested capital of Rs 12,00,00,00,000 are the same amount, so the business turns its capital exactly once, and that is only visible once the base is clean. On the inflated Rs 14,20,00,00,000 it appears to turn its capital 0.85 times and reads as a different sort of company.

What does the whole operation look like once it is run end to end?

Everything above lands in six lines and the table below is all of them. Two remarks before the arithmetic. Check one is already run inside the EBITDA line: it contains nothing from the land and nothing from the tooling holding. And depreciation and amortisationA yearly cost that spreads what a long-lived asset originally cost across all the years the business gets use out of it. is a real operating cost that none of these checks touches, so it stays exactly where the accounts put it.

Interest appears nowhere in that ladder, and its absence is the whole design of it. Notice too what the margins do when computed from the rupee lines rather than read off a label: Rs 2,40,00,00,000 on Rs 12,00,00,00,000 is a 20.0 per cent operating margin, and Rs 1,80,00,00,000 on the same revenue is 15.00 per cent after tax. Always take a margin from the amounts. A label can go stale while every amount beneath it stays correct.

Who actually runs these five checks, and what do they do with the answer?

A lender runs them to find out what is really available to service a loan. A borrower reporting comfortable profit that includes a one-off receipt and a slice belonging to a partner in a subsidiary has less coming in than the headline suggests. The covenant is usually written against a defined measure precisely so that both sides have agreed in advance which items are in it: that definition is check two and check four, run in a room with lawyers.

An equity analyst runs them because a comparison across companies is worthless unless every company in it was cleaned the same way. If one company's earnings per share has the minority taken out and another's does not, the two multiples are not measuring the same thing, and the difference will be read as a difference in the businesses.

A person buying into a small unlisted business does the same work with less paperwork and more at stake. The seller's last year shows a good profit. Some of it came from a machine sold off. Some is the owner paying themselves below a market wage, and no buyer will keep paying that. Some of the assets on the list are a car and a flat with nothing to do with the trade. Every one of those is one of the five checks, and running them is the difference between paying for a business and paying for a year.

A household does a smaller version whenever it works out what it can afford. Last year's income included a bonus that will not come again and rent from a room no longer let. The income left after both come out is the number a sensible commitment is sized against.

Where does normalising stop and steering the answer begin?

Every one of the five checks involves judgement. Whether a piece of land is genuinely surplus, whether a charge that has appeared three years running is really non-recurring, whether an owner's salary was above or below a market rate: none of these has a mechanical answer. A normalised base year is a set of choices, and a determined person can choose their way to almost any answer they want.

The line between cleaning and steering is not the size of the adjustments or how many there are; it is entirely whether each one was written down with its name, its amount, its direction and its reason. An adjustment recorded that way can be examined, argued with and reversed. The same adjustment folded quietly into a base figure cannot even be found.

So the discipline is disclosure rather than restraint. The log names every item and gives it a sign, an amount, and the reason in a sentence. A long list of well reasoned adjustments is more trustworthy than a single silent one. And the log changes what an argument can be about: it moves the disagreement from the answer, where it cannot be settled, to a specific line, where it can.

One more thing belongs in the log and almost nobody records it. When a check is run and finds nothing, write that down too. Sankalp Industrial Systems Limited's Year 0 has no non-recurring item, and the record says so in a line. A reader picking it up later can tell the difference between a check that found nothing and a check that was never run.

What must never be a step in this procedure?

Four things sit outside the judgement the last section described, and each of them is how a defensible procedure turns into an indefensible one.

Never adjust after seeing the valuation the adjustment produces. Once the answer is on the screen, every further change is being chosen against a target rather than against the accounts, and the person making it usually cannot tell the two apart from the inside. Run the five checks, write the log, close it, and only then build the forecast. Reopen a line only for a reason that would have applied before the answer existed.

Never call a recurring cost one-off because it is inconvenient. A restructuring charge in the third consecutive year is a cost of running that business, whatever it is labelled on the face of the statement. The test in check two is repetition, and it has no exception for items that spoil a margin.

Never adjust in one direction only. The instrument at the top permits exactly that: with a gain and a charge entered into the year and the removal of gains switched off, normalised EBITDA climbs above anything the year itself supports. A year offers two defensible readings, the one with everything left in and the one with every one-off taken out in both directions, and an honest pass lands between them. Outside that band, the figure about to be grown for five years is not a year this business ever had.

Never leave a check unrun and unrecorded. A step that found nothing produces a line saying so, and a reader who cannot find that line has no way of knowing the step ever happened.