Earnings Quality Checks You Can Run From the Statements
Earnings quality checks are a short list of comparisons run on published statements before a reported figure is trusted as the base for a forecast. Each one sets two numbers against each other that ought to move together: profit against operating cash flow, working capital against revenue, tax charged against tax paid, capital spending against depreciation. Each returns clear, look further, or cannot be run.
Every comparison below is built up separately, one at a time. The list is what assembly adds: eight named pairs, the order they run in, and a rule for what each is allowed to return. A comparison that is clear in isolation behaves differently when it is the fourth of eight and a score has started to be kept. Assembly is where a reader goes wrong.
Every figure belongs to one invented issuer, Sarvani Coatings Limited, a maker of decorative paints and industrial coatings, and to its published year three, the year ended 31 March. The instrument below runs all eight on those figures, and on any figures typed over them. The eight results never permit a statement about the company, and that restraint is the point rather than a limitation.
The eight checks, run on figures entered below
Every field opens on Sarvani Coatings Limited's published year three, so the instrument starts by reproducing the worked list further down, result for result. Changing any figure recomputes all eight. Nothing is stored anywhere: closing the tab discards the entries. A threshold is a rule and rules move, while arithmetic on two published figures does not. So every boundary in the instrument sits at equality or at zero.
| Check | The two numbers | What that reads | Result | Against the published record |
|---|---|---|---|---|
| 1 The adjusted figure, run both ways | ||||
| 2 Profit against operating cash flow | ||||
| 3 Working capital against revenue | ||||
| 4 Tax charged against tax paid | ||||
| 5 Other income against profit before tax | ||||
| 6 Capital spend against depreciation | ||||
| 7 The parts against the whole | ||||
| 8 The definitions against last year |
Eight checks, run on an ordinary set of statements. How many should be expected to come back clear?
What is a check, and why does every one of them need two numbers?
A check is a comparison. The comparison takes two figures out of the published statements, sets them side by side, and asks whether the relationship between them is the one to expect. There is no scoring, no weighting and no model.
A single number carries no information at all about quality. Every entry on this list exists to set two figures that ought to move together against each other. A household reports that its salary went up twelve per cent last year, which says nothing about whether it is better off, because what happened to the rent is unknown. Salary up twelve, rent up thirty, and something has been learned. Neither number said it alone.
Profit works this way too. Rs 278 crore of profit after tax in year three gives the size of the result and nothing about its texture. Set against the Rs 304 crore of cash the operations threw off in the same twelve months, the profit becomes a reading. The two measure the same trading under different timing rules, and one of them waits for the money.
Each row below is a pair, and the pair is chosen rather than convenient. Why a particular two numbers belong together is stated for every check. Arithmetic on the wrong pair still returns a confident answer, so that reason does more work than the arithmetic does.
Why does the result have three states and not two?
Two of them are obvious. Clear means the relationship is the one to expect, with no question to carry forward. Look further means it is not, and the question that raises gets written down. The third state keeps the exercise honest.
Cannot be run is a first-class result rather than a gap in the work. A checklist without it will always produce an answer, including on evidence that cannot support one. A two-state list cannot say it does not know, so when the data is missing it settles on the state that needs no evidence, which is clear. The result is a sheet of ticks, half meaning the check ran and passed and half meaning nothing at all, with nothing to tell them apart.
A tenant who has rented from one landlord once and has not yet moved out cannot tell whether deposits come back. Writing that down is useful, and the gap points to asking the previous tenant. A two-state form would force a tick one way or the other, and whichever was ticked would look, three months later, exactly like a finding.
Check eight is where that happens on Sarvani Coatings. The check compares this year's adjusted figure with last year's to see whether the same items were treated the same way. The record carries one year and no second to set beside it, so the check returns cannot be run, and that goes in the output word for word.
Why does the return rule have three states rather than two?
Check one: does the adjusted figure survive being run both ways?
For year three, management put Rs 452 crore forward as its adjusted EBITDA, lifting a Rs 6 crore restructuring charge onto the Rs 446 crore reported. The same notes to the accountsThe numbered explanations printed after the statements themselves, where an amount shown as one line on the face is broken into its parts. Their content is fixed by the accounting standards rather than by the reader. also disclose a Rs 4 crore provision write-backAn amount set aside in an earlier period and no longer needed, released back into the result. How and when a release happens is settled by the accounting standards. that went the other way and was not removed. Run the test in both directions and the figure is Rs 448 crore.
The presented figure and the both-ways figure belong together because the difference between them isolates the method from the arithmetic: both are correct additions, and the entire Rs 4 crore gap is a choice about which items count. Against the Rs 446 crore reported, Rs 4 crore is 0.90 per cent, and on the margin it moves 0.17 of a point. Small, until the presented margin of 18.72 per cent and the two-way margin of 18.55 per cent are each quoted as though they were the same kind of object.
Result: look further. Rs 452 crore is not wrong; it is arithmetically fine. The question is why the test was applied in the direction that lifts the figure rather than the one that lowers it, and that is answered in earlier reconciliations, a document this record does not carry.
Check two: does the profit turn into cash?
Profit after tax for year three was Rs 278 crore. Cash generated by the operations in the same twelve months was Rs 304 crore. The ratio is 1.09 times.
Profit and operating cash flow belong together because they measure the same period of trading under two different timing rules, so a persistent gap between them is a question about when things are being recorded rather than about how much was earned. Profit follows accrual accountingThe convention that revenue and costs are recorded in the period they belong to rather than the period the money moves. The accounting standards hold the rules for it., so a sale counts when it is made, while cash follows the bank. Last year's unpaid invoices come in while this year's go out, so in a settled business the two travel together.
Result: clear, for this year, and one observation establishes very little. A ratio above one for a single year fits a good year, a lucky collection month, or a business whose cash genuinely follows its profit. Note the limitation next to the result rather than upgrading the result to make up for it.
Check three: is working capital growing faster than revenue?
Working capital at the end of year three was Rs 335 crore: inventory of Rs 402 crore plus trade receivables of Rs 289 crore less trade payables of Rs 356 crore. A year earlier it was Rs 281 crore, a rise of Rs 54 crore or 19.2 per cent. Revenue over the same twelve months went from Rs 2,120 crore to Rs 2,415 crore, a rise of 13.9 per cent.
Working capital and revenue belong together because a trading cycle of a given length consumes a predictable share of revenue, so working capital growing faster than revenue means either the cycle lengthened or something inside it changed. The growth rates alone invite a shrug: nineteen against fourteen, and the business grew, so of course working capital grew. As a share of revenue the shrug gets harder, from 13.25 per cent at the end of year two to 13.87 per cent at the end of year three, a move of 0.62 of a point.
Result: look further, and the cycle days are where to look. Inventory is 112.5 days on the cost of materialsThe line covering what a maker paid for the raw inputs it consumed. Inventory and payables are both measured against it because both are stocks of materials rather than of sales., receivables 43.7 days on revenue, payables 99.6 days on the cost of materials again, and the cash cycle 56.6 days. The denominators differ on purpose, and mixing them is the commonest way this check gets run wrong.
Working capital up 19.2 per cent, revenue up 13.9 per cent, for the year ended 31 March. What is the next thing to look at?
Check four: does the tax charged match the tax paid?
Year three carries a tax charge of Rs 93 crore. Over profit before tax of Rs 371 crore the effective tax rateWhat the tax line works out to once it is set over the year result. An outcome rather than a rate anybody chooses, and the statutory rates feeding into it are published by the tax authority, never written here from memory. is 25.07 per cent. The cash flow statement shows Rs 88 crore handed over. On the same denominator that is 23.72 per cent. The gap is Rs 5 crore, or 1.35 points.
The tax charged and the tax paid belong together because the charge follows the accounting result and the payment follows the tax computation, so a gap between them is entirely ordinary and it is a widening gap across several years that is a question. A household knows this shape: the tax deducted from a salary over the year and the tax finally settled after filing are rarely the same number, and the difference is about timing.
Result: look further, with a note. The ordinary explanation is a passing timing differenceAn item counted in the accounting result in one period and in the tax computation in another, so the two figures separate for a while and then converge. The accounting standards and the tax law are what create one., which one year of data cannot distinguish from anything else. And at one decimal place, 25.1 less 23.7 prints a 1.4 point gap where the true gap is 1.35 points, so the second decimal is carried wherever the difference is the thing being reported.
Rs 93 crore charged and Rs 88 crore paid, in the year ended 31 March. Is that a problem?
Check five: how much of the profit came from outside the trading?
Other income for year three was Rs 38 crore against profit before tax of Rs 371 crore, a share of 10.24 per cent. Inside it sits an insurance claim of Rs 9 crore disclosed in the notes, leaving Rs 29 crore as the ordinary run of that line.
Other income and profit before tax belong together because other income is where amounts unconnected with the trading collect, so its share of the result measures how much of the profit arrived from somewhere other than making and selling paint. A tea stall makes money from tea. If a third of last month's income was a one-time payment from a cousin, the stall did not get better, and next month's plan should not assume the cousin.
Result: look further on the claim, clear on the remainder. The Rs 9 crore claim is 2.43 per cent of profit before tax. The question it raises is not what the claim means but how often an amount of its kind turns up in that line, and answering that needs more years than this record carries. The Rs 29 crore remainder raises nothing by itself, though it would if it ever became the reason the profit grew.
Check six: is the capital spending in scale with the asset base?
Capital spending in year three, from the investing section of the cash flow statement, was Rs 186 crore. Depreciation and amortisation in the same year was Rs 92 crore. The ratio is 2.02 times.
Capital spending and depreciation belong together because depreciation is a rough scale for the asset base already in place, so the ratio between them says whether the business is spending like one that is growing or one that is holding steady. At about one times a business is replacing what wears out. At two times it is adding something. Well below one, sustained, is a business living off equipment it is not replacing. The shortfall shows up in nobody's profit for several years and then all at once.
Result: clear, with a note. At the year end the balance sheet carries Rs 118 crore of capital work in progressSpending on an asset not yet finished or in use, parked on its own balance sheet line until it is. The accounting standards fix what qualifies and when it moves off that line., a coatings line waiting to be commissioned. The unfinished line explains spending at twice depreciation, on the face of the statements rather than hidden. Do not confuse it with the Rs 118 crore of free cash flow the same year produced: two unrelated amounts that happen to be equal, exactly the coincidence that survives into somebody else's note as a fact.
Check seven: do the parts add up to the whole?
Quarter by quarter, revenue arrived as Rs 590 crore, Rs 545 crore, Rs 700 crore and Rs 580 crore, adding to Rs 2,415 crore, the published year. Quarterly EBITDA of Rs 106 crore, Rs 88 crore, Rs 145 crore and Rs 107 crore adds to the published Rs 446 crore. The two segments, decorative at Rs 1,811 crore and industrial at Rs 604 crore, add to Rs 2,415 crore once more.
The summation check belongs on the list because it costs a minute, it catches a restatementA published figure reissued with a different value, usually because a classification changed. The disclosure rules, not the reader, settle whether one must be announced and how. nobody announced, and it is the only check here that can fail mechanically. When it fails, everything else becomes unreliable at once, because every other check is arithmetic on figures that have just been shown not to agree.
Result: clear, on all three sums, with a residual of Rs 0 crore in each. A clean tie removes one whole class of explanation for the four look-further results, before any real time has been spent on them.
The four quarters sum to the published year and the segments sum to revenue. What has that established?
Check eight: are the definitions the same as last year's?
The definitions check has no arithmetic in it. The adjusted figure presented this year is set beside last year's, and what was added back on each occasion is compared. If the restructuring charge came out this year, did an equivalent charge come out last year, and if a favourable item was left in this year, was one left in last year too.
The definitions check is the one most often skipped, and the reason is not that it is hard but that it needs a document that is not already open, rather than any new calculation. Everything else runs on the documents already to hand. Running this one means going back to a filing from twelve months ago, finding the reconciliation table and reading it properly. Fifteen minutes of retrieval is exactly the friction that stops a check being run.
Result on this record: cannot be run. Sarvani Coatings Limited discloses an adjusted figure for one year and there is no second reconciliation to set beside it. Write cannot be run, name the document that would change that, and move on without upgrading the result.
One year of adjusted figures is available and no earlier reconciliation. What does the definitions check return?
In what order do they run, and why that order?
Cheapest and most mechanical first. Cost order is the whole rule, and the rule is worth more than it sounds.
Five of the eight run on the face of this year's statements. Two need the notes and the comparative column, in the same document but requiring a search. One needs a document from a different year altogether.
The expensive checks consume the attention and the hour, so a reader who starts with them very often never reaches the cheap ones, and the free results end up never collected. Start household paperwork with the insurance renewal that needs three phone calls and the evening goes, with the four jobs that needed two minutes each still sitting there.
The whole list, run on one year of statements
End to end, on Sarvani Coatings Limited's published year three. A check whose inputs cannot be located is a check that will not be run twice, so the third column names where each of the two numbers is found.
| Check | The two numbers | Where each is found | Result |
|---|---|---|---|
| 7 The parts against the whole | Rs 2,415 cr Rs 2,415 cr | The four quarterly filings; the segment note in the annual accounts | clear |
| 2 Profit against operating cash flow | Rs 278 cr Rs 304 cr | Profit and loss, last line; cash flow statement, operating section total | clear, one year only |
| 5 Other income against profit before tax | Rs 38 cr Rs 371 cr | Profit and loss, both on the face; the claim is sized in the notes | look further |
| 4 Tax charged against tax paid | Rs 93 cr Rs 88 cr | Profit and loss, tax line; cash flow statement, tax paid line | look further |
| 6 Capital spend against depreciation | Rs 186 cr Rs 92 cr | Cash flow statement, investing section; profit and loss, depreciation line | clear |
| 3 Working capital against revenue | Rs 335 cr Rs 2,415 cr | Balance sheet, both columns, three lines each; profit and loss, first line | look further |
| 1 The adjusted figure, run both ways | Rs 452 cr Rs 448 cr | The reconciliation in the results release; both items sized in the notes | look further |
| 8 The definitions against last year | one year no second | Last year results release, which this record does not carry | cannot be run |
| Eight checks | 3 / 4 / 1 | Seven of the eight runnable on this record | see below |
Row two has to be read carefully. Profit against cash returned clear at 1.09 times, and the words one year only stay in the result column rather than dropping off once the row is counted. A ratio built to be read as a trend, observed once, is a pass with a limitation attached, and the limitation travels into whatever gets written next. A reader who logs the row as look further until a second year exists is running the same list honestly. Logging it clear and letting the note fall off is not.
The tally is three clear, four look further and one that cannot be run, and that is a completely ordinary outcome on a completely ordinary set of statements. Nothing about it is alarming, nothing about it is reassuring, and it is not a score out of eight.
Four more years of comparable documents become available. Which checks actually improve?
How much of the list can actually be run?
The control sets how many years of comparable documents are held. Some rows change state, some get a stronger reading without changing state, and some never move at all. Selecting a row has the readout explain that one.
What happens to a look further, and what does the count mean?
A look further produces three things and nothing else. A question, written as a sentence. A document to look in. A date by which that document should be open to reading. One line each.
The four look further results on this record give four of those triples, set out in the figure below. Each names a question, the document that answers it, and the filing by which that document should become available. Three of the four wait on the next annual filing and one on the next half year.
Four look-further results is a statement about how much work is outstanding, and it is not a score, not a grade, and not a finding about Sarvani Coatings Limited. A four means four questions to work through, not that the earnings are worse than a two would have implied. A high count very often means the opposite. An issuer that discloses more gives these checks more to catch on, and one that discloses less returns clear more often because there is nothing there to compare.
The failure, and what it costs
An analyst runs the list, counts four look-further results out of eight, and writes that the earnings are of poor quality. The conclusion feels rigorous. Eight checks, half of them failed, a defensible number.
Every one of those four has an ordinary explanation sitting in the same set of documents. A working capital build in a business whose revenue grew 13.9 per cent. A timing difference between a tax charge and a tax payment. An insurance claim disclosed in the notes rather than hidden. And an adjustment policy that a great many issuers apply without anybody thinking about it.
The cost is worse than a wrong answer, and this is the part worth sitting with. A wrong answer can be corrected. A conclusion stops the work. An analyst holding the view that the earnings are poor does not write the four questions, does not open the four documents, does not put four dates in the diary, and arrives at the next results release with nothing to check and a view they now have to defend. Six months of possible evidence goes past uncollected.
The fix is mechanical. Each look further produces a question, a document and a date. The number of them is a measure of the outstanding work. Nothing else comes out of the count.
Can the count improve while the statements get thinner?
The count can improve, and the instrument above shows it in one click. Loading the second set of figures, the one labelled an issuer that discloses less, changes four fields, and none is a trading figure: no charge is added back, so no adjusted figure is presented; the provisions note breaks out nothing; no one-off inside other income is sized; and tax paid equals the charge. Revenue, profit before tax, the quarters and the balance sheet are untouched.
Three checks move from look further to clear, the tally goes from three clear to six, and the business is exactly as well understood as before. The count improved because there was less to compare, which is the sharpest available demonstration that the count measures the documents rather than the trading.
So a reader who treats four look-further results as a grade has the sign backwards: the fuller set of accounts, the one that gave these pairs something to catch on, scores worse. The two lists of questions matter, not the two tallies. The thin issuer produces one question and the fuller issuer four, and four questions that can be gone and answered beat three extra ticks that cannot be interrogated.
Run the list on an issuer that presents no adjusted figure, sizes no one-off inside other income, and shows tax paid equal to the charge. What happens to the tally?
Who actually runs this list, and what do they do with it?
A lender runs it before renewing a working capital line, and the rows they care about are not an equity reader's. Check three is the whole conversation: working capital at 13.87 per cent of revenue against 13.25 per cent, and a cash cycle of 56.6 days, sets how much facility the business needs next year. Check two tells them whether the profit they are lending against arrives as money.
An equity analyst runs it to decide which reported figure becomes the base of a forecast. Check one moves the starting EBITDA by Rs 4 crore. Check six tells them whether Rs 186 crore of spending is a one-off build or the running rate. The answer changes every year of the projection rather than only the first.
A household does the same without the vocabulary: before lending a cousin money for the shop, the questions are whether last year takings turned into cash, whether the stock in the back is growing faster than the sales, and whether the tax he says he paid matches the tax he says he owed. Three checks, two numbers each, no verdict, just a shorter list of things to ask him.
All three uses share the same shape: the list decides what goes into the next stage of work, and none of the three uses it to reach an opinion about the people who published the figures.
Four checks returned look further. What has been established about the issuer?
Why does not one of these eight checks quote a rule?
Eight checks, and not a single percentage limit, filing window, tax rate or stated obligation among them. The absence is not an omission. Every one of those moves, and a checklist that hard-codes last year version of one keeps returning a confident result long after the result stopped being true. Arithmetic on two published figures does not go stale. A quoted threshold does.
So the checks stop where a rule would start and hand the question on. How a provision release, or a timing difference between the charge and the payment, is measured and shown belongs to the accounting standards, held by the Institute of Chartered Accountants of India at icai.org, inside a statutory frame kept by the Ministry of Corporate Affairs at mca.gov.in. The securities regulator at sebi.gov.in settles what a listed issuer must publish and what a research writer may set next to a number. The filings checks seven and eight run against come down from the exchange archives, at nseindia.com and bseindia.com, each lodged copy stamped with when it went up. The stamp is lifted off the lodged document rather than off anything summarising it.
Where do these eight results come from, and what is not behind any of them?
About the numbers. The year three statements of Sarvani Coatings Limited, ended 31 March, and the notes the two disclosed items sit in, were built so that every check can be rerun by hand, or through the instrument above, and land on the same eight results. Where a check meets how an item is measured rather than what it equals, the body that settles the measurement is named. No result was derived from a rounded percentage: every one was computed from whole rupees and rounded only for display.
Which check hands off where
| Where a check stops | Who settles what is beyond it | Site | Read on |
|---|---|---|---|
| Check four meets the difference between a tax charge and a tax payment; check one meets a provision that was released. How either is measured and disclosed under Ind AS is decided here. | Institute of Chartered Accountants of India | icai.org | consulted 28 August 2026 |
| The statutory frame the annual accounts, the notes and the segment disclosure that checks five, seven and eight read from are prepared inside. | Ministry of Corporate Affairs | mca.gov.in | consulted 28 August 2026 |
| What a listed issuer must publish, and what a research writer may set beside a number they have computed. No period and no percentage limit is written above. | Securities and Exchange Board of India | sebi.gov.in | consulted 28 August 2026 |
| One of the two places the filed result and its attachments can actually be downloaded, which is where checks seven and eight get run. | National Stock Exchange of India | nseindia.com | consulted 28 August 2026 |
| The second lodging of the same result, worth knowing about when check eight needs last year document and one archive is thin. | BSE Limited, the Bombay Stock Exchange | bseindia.com | consulted 28 August 2026 |
Sarvani Coatings Limited, Thottam Chemicals Limited, Kesaria Surface Solutions Limited, Nandivarman Paints Limited, Meghna Iyer and Ravindra Setlur are invented.
Educational material. Not advice on any investment, tax, budget or market position.
