The forensic scorecard
Net income is an opinion. Cash is a fact. One of these filers cannot be asked the question at all.
Cash conversion is the first of the brief’s six pillars to be built, the balance sheet is the second, revenue quality is the third and governance is the fourth. Nine exhibits. Three of them read one annual report that had been opened once for megawatts alone, and the ninth is the one governance question a machine harvest can answer. The first runs two measures of one idea and reports them side by side rather than averaging them, because they can disagree and the disagreement is worth more than either verdict alone. The bands are published on the methodology page and read from the same file that applies them here.
Exhibit 1
One pillar, five filers, three currencies and no conversion between them
One row per filer, on that filer's own most recently filed year. Three of these companies close on 31 March and two on 31 December, so the year is printed on every row rather than assumed by a column heading. Both measures are ratios, so nothing here depends on the currency the accounts were filed in.
| Filer | Year filed | Operating cash to profit after taxtimes | Sloan accrual ratioper cent of total assets |
|---|---|---|---|
| FY2026 | Profit after tax is not positive in this period, and a ratio against it would change sign without the cash changing. | 4.8%within | |
| Equinix, Inc.Accounts in USD | FY2025 | 2.90xwithin | No balance sheet is filed at this date. An annual report prints its statement of financial position one year behind the cash flow filed beside it. |
| Digital Realty Trust, Inc.Accounts in USD | FY2025 | 1.84xwithin | No balance sheet is filed at this date. An annual report prints its statement of financial position one year behind the cash flow filed beside it. |
| Wipro LimitedAccounts in INR | FY2026 | 1.13xwithin | 1.2%within |
| Infosys LimitedAccounts in USD | FY2026 | 1.22xwithin | -4.5%within |
The empty cells are empty for different reasons and the page keeps them apart. The Indian operator’s group made a loss in this period, so a ratio against profit would change sign without the cash changing, and the measure refuses rather than returning a negative that would sort like a grade. The two US filers are missing a denominator instead: an annual report prints its statement of financial position one year behind the cash flow filed beside it, so the most recent year has cash on both sides and no balance sheet to divide by. One absence is about the company. The other is about the document.
Source: Harvested from the filings store one concept at a time, with the filing that served each figure recorded against it. Filings are 10-K for the two US filers and 20-F for the three that report to the SEC as foreign private issuers.
Exhibit 2
The group scores 35 times on this measure in the year its profit almost disappeared, and refuses it the year after
Two documents measuring one business, on the years both filed. The arm files restated statements into a draft prospectus in rupees millions; the parent files a 20-F in absolute rupees. Neither figure is converted and neither needs to be, because both measures divide the scale out. The stub quarter is excluded, since the parent files no quarter to set beside it.
| Filer | Year filed | Operating cash to profit after taxtimes | Sloan accrual ratioper cent of total assets |
|---|---|---|---|
| Sify Infinit Spaces LimitedAccounts in INR, millions | FY2023, standalone | 3.86xwithin | 22.8%within |
| FY2023 | 12.03xwithin | 11.0%within | |
| Sify Infinit Spaces LimitedAccounts in INR, millions | FY2024, standalone | 2.26xwithin | 21.2%within |
| FY2024 | 35.12xwithin | 9.4%within | |
| Sify Infinit Spaces LimitedAccounts in INR, millions | FY2025, consolidated | 4.33xwithin | 3.1%within |
| FY2025 | Profit after tax is not positive in this period, and a ratio against it would change sign without the cash changing. | 3.5%within |
Read the parent’s row and the measure improves from 12.03 to 35.12 times, then stops resolving. Nothing improved. Profit after tax went from 1532 million rupees in FY2021 to a loss of 1366 million in FY2026, while operating cash stayed in the thousands of millions. The quotient climbed on a shrinking denominator and then hit a negative one. A ratio against profit is at its least informative exactly where the profit line is in trouble, which is why the measure refuses a non-positive denominator rather than printing the number.
The subsidiary underneath it is the steady one, turning profit into between 2.26 and 4.33 times as much operating cash, above the 0.8 flag in every year. Its accrual ratio runs far higher than the parent’s in the two standalone years, which is the build showing through rather than the earnings: the measure nets investing cash back into its numerator, and this company is spending. The two documents are never reconciled to each other, and the gap between the parent’s segment note and the arm’s own accounts is separately a constant 87.7 million rupees in every year they overlap.
Source: The arm from Sify Infinit Spaces DRHP, Restated Consolidated Statement of Cash Flow. Capital expenditure is the purchase of property, plant and equipment line. Amounts paid to acquire right of use assets are reported separately in the same statement and are held separately here. Tax paid and capital expenditure are outflows in the filing and are recorded as positive magnitudes. The parent from its 20-F, as served by the filings store. Printed page 355.
Exhibit 3
The document never says whether a lease is a borrowing. Its own net debt line answers it in every period
Rupees millions and times earnings, for the operator whose statements are read page by page. Net debt as the issuer publishes it, rebuilt from the balance sheet lines beside it, and the leverage ratio on both readings of the lease liability. Source at printed page 353.
| Period | Net debtas published | Rebuiltleases counted | Net debt to EBITDAas published | On the printed wordingleases removed |
|---|---|---|---|---|
| FY2023 | 16,007 | 16,007exact | 3.88x | 3.31x |
| FY2024 | 21,503 | 21,503exact | 4.62x | 4.00x |
| FY2025 | 23,236 | 23,236exact | 3.66x | 3.18x |
| Q1 FY2026annualised by the issuer | 27,390 | 27,390exact | 3.87x | 3.03x |
One period is published on a different basis from the earnings printed for it. Dividing the published net debt by the published ratio recovers the earnings the issuer used, and for Q1 FY2026 that figure is 4.00 times the earnings the same period reports. The quarter is annualised for this ratio. The same document reports the same quarter’s return on capital without annualising it, and labels neither.
The return on capital formula this issuer prints defines capital employed as net worth plus total borrowings less cash, and never says whether a lease liability counts as a borrowing. That question decides the answer: the published returns only rebuild on the reading where it does, which is an inference drawn backwards from a result. The net debt line settles it forwards. In all 4 filed periods the issuer’s own published net debt equals borrowings plus lease liabilities less cash, to the paisa. The convention is written nowhere and provable everywhere.
It is also the stricter of the two. Read the printed wording literally and leverage falls to 3.31 times from 3.88 in the first filed period. The issuer took the reading that flatters it less, here and on the return on capital, and states that choice in neither place.
One more thing falls out of the same arithmetic. Dividing published net debt by the published ratio recovers the earnings figure behind it, and for Q1 FY2026 that is 4.00 times the earnings the same period reports, so the quarter is annualised for this ratio. The same document reports that quarter’s return on capital unannualised. Two ratios, one period, opposite conventions, and neither is labelled.
Source: Sify Infinit Spaces DRHP, Restated Consolidated Statement of Assets and Liabilities. Net worth is equity share capital plus other equity. Borrowings are the non current and current borrowing lines added. Lease liabilities are the non current and current lease lines added, held apart from borrowings because which of the two belongs inside total borrowings is the question the printed formula leaves open. Cash is the cash and bank balances line, excluding other bank balances. Printed page 353.
Exhibit 4
Sify Technologies Limited refuses least of the three, and the reason it can is that nobody asks it much
Questions pressed on unit economics and how many were refused, over 2024-01-01 to 2026-07-31. Bars are on one scale, so the width is the asking and the filled part is the refusing. Only a declined or deflected answer counts as a refusal; a partial answer is an answer.
0.43 unit economics questions a call, across 35 calls
0.81 unit economics questions a call, across 42 calls
0.48 unit economics questions a call, across 42 calls
2 of the 20 refusals named somewhere the figure could be found. That second measure carries a limit worth stating beside it rather than under it: it is coded from what management said out loud, so a refusal naming no source is not evidence the figure is unpublished. Both global operators publish quarterly supplements that a call answer may simply not mention.
The expected shape of this measure is an Indian operator that says less than its global peers. It runs the other way. Sify Technologies Limited refuses 2 of 15 questions on its own unit economics. Equinix, Inc. refuses 9 of 20.
The rate alone would still mislead, which is why the denominator is the bar rather than a footnote. Digital Realty Trust, Inc. is asked 0.81 unit economics questions a call against 0.43. A company nobody presses has little to refuse. The sharpest single instance runs against the thesis too: asked for realisation per megawatt, the Indian operator declined and then described the reverse working, which is the calculation this analysis performs. The company confirmed the method in its own words while declining the number.
Source: Earnings call transcripts across 119 calls. The two topic families that bear directly on unit economics. The same two are used for every company, so the denominators are comparable. Calendar dates, not fiscal labels. The fiscal labels in this source are not internally consistent across filers, so the window is defined on the date the call happened.
Exhibit 5
A year reported up 74 per cent ends at a rate that annualises 18 per cent below it
E2E Networks Limited, FY2024-25. The top three bars are crore of revenue on one scale. The bar beneath them is the profit before tax, split into what the operations produced and what arrived as other income, in INR lakh. Both units are as the report prints them, two pages apart, and neither is converted into the other.
Revenue rose 73.57 per cent over the year. The rate it ended on rose 2.75 per cent, and annualises 29.6 crore below the year it just finished, or 18.0 per cent of it. Twelve times a month is a run rate rather than a forecast; the point is that both figures describe the same twelve months. Printed page 26.
- From operations
- 2,336 lakh, 37%
- Other income
- 3,943 lakh, 63% Printed pages 24 and 91
Other income is 1.69 times what the operations produced before tax, against 163 lakh the year before. The operating result is thin against earnings before interest, tax and depreciation of 9,666 because depreciation took 6,008, which is what a year of buying accelerators looks like once it reaches the income statement. What the other income is, the report does not say, and it is not guessed at here.
This is the coverage name with no capacity to measure. Its report states no megawatt and no kilowatt in 148 pages, and it is a tenant in someone else’s hall rather than an operator, so the question this analysis usually asks cannot be put to it. Two questions can, and both come out of figures it prints itself.
The first is where in the year the growth happened. The only operating rate the company publishes is monthly recurring revenue, in its own words: “Compared to Rs. 10.90 Crores in March 2024, the monthly recurring revenue increased to Rs. 11.2 Crores in March 2025.” That is 2.75 per cent against a year reported up 73.57. A year of that shape has its growth behind it rather than in front.
The second is where the profit came from. Other income of 3,942.68 is 1.69 times the 2,336.46 the business made before tax, and 63 per cent of the profit before tax. The report gives the figure and never its composition. The same summary table is printed twice, in the directors’ report and again in the management discussion, with identical figures and no explanation in either.
Source: E2E Networks Limited annual report FY2024-25. The financial performance summary is printed twice, in the Directors' Report and again in the management discussion, with identical figures. The monthly recurring revenue sits in the Directors' Report narrative. Printed page 24.
Exhibit 6
The target is 250 megawatts. The contracts signed for capital spending buy 0.20.
Three lines from one annual report, in millions of rupees, none of them converted. The megawatt band underneath divides the contracted capital by the sector build cost of 60 to 70 crore per megawatt, which is a sector figure rather than this filer's, because no operator in this coverage publishes its own.
Millions of rupees, all three from the year ended 31 March 2026. The disputed tax is 2.13 times the contracted capital and the overdue balances are 7.59 times it. Printed pages 264 and 368 for the commitment, and 174 for the auditor’s paragraph.
- Targeted by FY 2029-30
- 250 MW
- Commissioned and live
- 24 MW
- Contracted
- 0.17 to 0.20 MW
The bar is the contracted megawatts against the target, and it is drawn at a floor width so that it renders at all. Printed page 73 for the target.
Across 433 printed pages the accounts carry no segment disclosure. The terms searched were Ind AS 108, reportable segment, operating segment, segment information, chief operating decision maker. So the business the report calls “Our data centre business is the most consequential strategic decision we have made in a generation” on printed page 72 has no revenue, no margin and no asset base a reader can separate from the engineering business that funds it. Every megawatt figure above is management commentary; every rupee figure is audited.
The capacity target and the campus megawatts are management commentary, printed in the management discussion. The three bars are audited. Setting them beside each other is the whole measure: the only line in either set of accounts recording contracts for future capital spending is 2.13 times smaller than the tax the company is disputing, and 7.59 times smaller than the balances its auditor drew attention to as substantially overdue and carrying no impairment provision.
The commitment did rise, from 29.35 million a year earlier to 118.05, a fourfold rise on a base small enough that fourfold changes nothing. The smallest campus the report names, Kolkata at 8 MW, costs 41 times the whole commitment at the low end of the sector rate.
The struck off disclosure on the same filer runs to 1 counterparty, a trade payables balance of 0.02 million with Pyrotech Electronics Private Ltd, against eleven on the other annual report read here. The consolidated note numbers two clauses (ii). The first tables that balance. The second reads, in full, “The Group do not have any transactions with struck off companies.” Printed page 381.
Source: Techno Electric & Engineering Company Limited annual report FY2025-26, note 38 to the standalone financial statements and note 37 to the consolidated financial statements, contingent liabilities and commitments. The auditor's paragraph is quoted from the independent auditor's report on the standalone financial statements. Printed page 368.
Exhibit 7
A liability moved from payables to borrowings, and the cash flow statement records nothing moving
Two payment ranges on one scale of days, then one reported balance split by how it arrived. Millions of rupees, from the standalone statements, converted nowhere. The disclosure exists because amendments to Ind AS 7 and Ind AS 107 were notified partway through the year the report covers.
Days after the invoice date, on one scale. The financed terms start 60 days later than the comparable ones and end 90 days later, so the company pays at roughly twice the remove it does outside the arrangement. Interest runs at SOFR plus 70bps, against collateral. Printed page 243.
- Moved from trade payables, no cash
- 194.59, 97%
- Actually borrowed
- 5.31 Printed page 267
The arrangement is 1.66 times the 118.05 the company has contracted to spend on capital account. Two things about it point opposite ways and both belong here. Placing the liability under borrowings rather than leaving it among payables is the stricter of the two treatments available, and the company chose it. Recording the move as a non cash transfer is also correct, and its effect is that a doubling of the time taken to pay leaves no mark on either half of the cash flow statement. Against the payables it came out of the amount is small, at 1.71 per cent of the total. The disclosure exists at all because the amendments requiring it were notified partway through the year, on printed page 217.
Supplier finance pays a vendor early and collects from the buyer later. The buyer’s own terms lengthen and the obligation stops being a trade payable. Here the terms roughly double, from 60 to 90 days to 120 to 180, and 97 per cent of everything the standalone reports as current borrowings arrived by that move rather than by borrowing.
The company classified it under borrowings rather than leaving it among payables, which is the stricter of the two treatments and cuts against the reading above. Both are on the page for that reason. What the treatment cannot do is put the lengthening into the cash flow statement: the transfer is non cash, so neither operating nor financing activities record it, and the same figure covers 60 per cent of the rise in consolidated current borrowings across the year.
Source: Techno Electric & Engineering Company Limited annual report FY2025-26, note 21 to the standalone financial statements at printed 243 and note 19 to the consolidated at printed 349, with the non cash reclassification recorded in the net debt reconciliations at printed 267 and 369
Exhibit 8
One note places the whole micro and small enterprise balance past its due date. The next reports nil interest, five times over.
Millions of rupees from the standalone statements, converted nowhere. The ageing columns run from the due date of payment rather than from the invoice date, so every column except the first holds amounts already late.
- Not due
- 2,127.53
- Less than 1 year
- 8,806.12
- of which micro and small 403.83
- 1 to 2 years
- 229.80
- 2 to 3 years
- 39.10
- More than 3 years
- 159.07
The columns run from the due date of payment, so everything outside the first is already late: 81.3 per cent of the balance, against 86.5 a year earlier. The whole micro and small enterprise balance of 403.83 sits in an overdue column and 0.00 in the column for amounts not yet due, in both years, while the balance rose 83.7 per cent. Printed page 244.
What the filing reports under the Act, one page later
| Clause | What it requires | Reported |
|---|---|---|
| (a) | Principal and the interest due thereon remaining unpaid | nil |
| (b) | Interest paid under the Act along with payments made beyond the appointed day | nil |
| (c) | Interest due and payable for the period of delay | nil |
| (d) | Interest accrued and remaining unpaid at the year end | nil |
| (e) | Further interest remaining due and payable in succeeding years | nil |
5 of 5 report nothing, in both years. Printed page 245.
The shortest overdue bucket the table prints covers everything up to a year, and the appointed day under the Act is 45 days. The bucket is therefore coarser than the statute, and cannot show that every rupee passed the appointed day. What the five clauses report is that none of it did. Two further things belong beside it. The filing says the balance was identified as follows: Dues to Micro and Small Enterprises have been determined to the extent such parties have been identified on the basis of information collected by the Management. So its completeness rests on what suppliers told the company. And letters of credit, secured against the plant, the fixed deposits and the receivables, now stand behind 5,343.84 of these payables, 47 per cent of the total and up 75 per cent on the year, which is a third way of paying later alongside the two above.
Interest on a payment made after the appointed day is automatic under the Act rather than something a supplier has to claim. The filing reports 5 of 5 clauses as nothing, in both years, one printed page after the table that placed 403.83 million past its due date. This is the second time in one document that a clause tables a balance and a neighbouring clause says there is none to table.
What the table cannot settle is stated with it rather than left out. The shortest overdue bucket runs to a year and the appointed day is forty five days, so the balance being past its due date is not the same as every rupee of it having passed the day the statute names. The five clauses are what report that none of it did.
Source: Techno Electric & Engineering Company Limited annual report FY2025-26, note 22 to the standalone financial statements, trade payables ageing at printed 244 and the disclosure under the Micro, Small and Medium Enterprises Development Act 2006 at printed 245
Exhibit 9
3 of these 5 filers never went back and changed a published figure. One changed 39 per cent of what it published.
One cell per annual data point harvested from that filer, and the track is as long as the filer was asked. Filled cells are periods the filings store reports as restated. Rates over unequal denominators, which is why both are drawn.
Assets (IFRS), Cash flows from used in investing activities, Profit loss (IFRS) · 2023-03-31, 2024-03-31, 2025-03-31
Assets, Net cash provided by used in investing activities, Net income loss, revenue · 2022-12-31, 2023-12-31
Each track is one harvested annual point per cell, and the track is as long as the filer was asked. Equinix, Inc. has the longest track at 76 points and restated 7 of them. 3 of 5 filers restated nothing at all.
Restatement is the one governance signal in the brief that a machine harvest can answer without a document being opened by hand, because the store records which periods a later filing changed. Three of the five changed nothing across 100 harvested points between them.
The rate has to be read against the length of the bar and not on its own. The operator this coverage is built around is asked for 3 concepts and has 18 chances to restate; Equinix, Inc. is asked for 9 and has 76. A count alone would put the second at the top and a rate alone would put the first there, so both are on the page. This is the second measure here to need that correction: the refusal rates in Exhibit 4 found that the company refusing least was the one asked least.
Source: Restated periods as the filings store reports them per concept, promoted from data/raw/filings into each company file so the count is validated at build.
What is not here
Six pillars against ten subjects, and 22 of the 60 cells filled
Every pillar against every subject a filing has been read or harvested for. A filled cell is a measure that computes and renders. The empty ones are the point.
22 / 60
- Sify Infinit Spaces
- Sify Technologies
- Anant Raj
- Techno Electric
- E2E Networks
- Netweb
- Equinix
- Digital Realty
- Infosys
- Wipro
The three names with no statements read cannot be asked it. Operating cash and profit after tax are what it needs, and neither is filed in a form this coverage holds for Techno Electric, E2E or Netweb.
The five harvested filers carry total assets and no borrowings or lease liability split, and that harvest was attempted once and abandoned because the store returns maturity schedule rows that look like balances.
Restatement is the one governance question a machine harvest can answer, and it is now asked of all five harvested filers. Everything else on this pillar still needs a document opened by hand: auditor and officer turnover, holding and pledging, and audit qualifications are unread for the five. Neither E2E nor Netweb has a filing read for that purpose at all.
Concentration and growth only. Pricing, contract duration and churn are unread for every subject, and one single customer revenue line running 2017 to 2026 is filed and unpulled.
One clean answer and no measure. Segmentation is not like for like across the set: two subjects are pure plays, two segment by industry vertical, one by service line, and one discloses no segments at all.
Terminal rather than pending. Every figure here comes from a filed document, no market data source is in the repository, and a draft prospectus carries no price band. A half sourced multiple is worse than no multiple.