How to Build a Stock Screen in India

Run the most popular screen in India — ROE above 15%, debt-to-equity below 0.5, sales growth above 15% — and you get back the same 200-odd names everyone else gets back. The screen worked. It just didn't help. A filter that returns the consensus list has told you nothing the market doesn't already price.

That is the quiet failure mode of screening: not that it returns garbage, but that it returns crowded quality at full price. And it has a second, deeper hole. A screen reads the financial statements — numbers that are, by definition, history. It cannot see whether the management running the business does what it said it would, which in India is most of the thesis. The pattern Inve sees, tracking management guidance across 1,500-plus listed companies, is that talk and delivery diverge far more than most investors assume: fewer than one in five guidance items has actually been delivered as stated, a large share is still open, and a meaningful number were simply never mentioned again on any later call. (This is a read on how management has communicated over roughly the last two years of transcripts — not a lifetime verdict — and many commitments are still in flight; Inve data.) No price-and-ratio screen will ever surface that. This piece is about building a screen that does narrow — and being honest about what it structurally can't catch.

(Companies named below are illustrations of the method, not views on the stock.)

Why do most stock screens fail?

A screen is a sieve. Its only job is to take a universe of 4,000-plus listed companies and hand you a shortlist small enough to actually research. It fails in three predictable ways.

It fails when it is too loose — twelve filters, each set generously, and you get back 600 names. That isn't a shortlist, it's the market with a haircut. The point of a screen is brutal subtraction, not a gentle trim.

It fails when it is too crowded — the textbook quality screen (high ROE, low debt, steady growth) returns names every fund and every finfluencer already owns. The companies are genuinely good. The prices have long since stopped being interesting. You've screened for quality and forgotten that you also have to pay for it.

And it fails — most importantly — when it is backward-looking and blind to delivery. Every line in a screener comes from a filed statement: last year's ROE, last quarter's sales, the reported debt. All of it is the rear-view mirror. None of it tells you whether the management guiding "we'll be operating-cash-flow positive by next quarter" has any history of hitting what it guides. That is forward information, it lives in the earnings calls, and no ratio captures it.

A screen that works fixes all three: it subtracts hard, it deliberately looks where the crowd isn't, and it treats management's delivery record as a filter, not an afterthought.

What are the three filters every screen needs?

Think of a working screen as three sieves stacked in order — quality, value, delivery. Each one removes a different kind of company, and the order matters, because you want to spend your scarce research time only on names that survive all three.

Quality — is this a business worth owning at any price? This is where you keep the ratios, but read in pairs, never alone. Return on capital employed (ROCE) tells you how productively the whole business uses its capital; return on equity (ROE) tells you the return to shareholders. Read together, a wide gap between them means leverage is doing the work, not the business. (We unpack that fully in ROCE vs ROE: what the gap reveals.) Pair revenue growth with operating cash flow — profit that never becomes cash is an accounting opinion, not money (see quality of earnings: profit vs cash flow). Pair margin with debt — a thin-margin, high-debt business has no shock absorber.

That cash-flow pairing is the one most screens skip, and it is the one that bites. Take Kaynes Technology, the electronics-manufacturing name. Revenue more than tripled from ₹1,126 crore in FY23 to ₹3,626 crore in FY26, and reported profit rose every year — ₹95 crore, ₹183 crore, ₹293 crore, ₹364 crore (Inve data, 2026). A ROE-and-growth screen waves it straight through. Now pair the profit with the cash: operating cash flow ran negative ₹42 crore, then positive ₹70 crore, then negative ₹82 crore, then negative ₹600 crore (company filings, consolidated) — negative in three of those four years, and in FY26 the business burned ₹600 crore of operating cash against ₹364 crore of accounting profit. The profit was real on paper; it just never turned into money in the bank, because working capital swallowed it. A growth manufacturer that consumes cash to run, before a single rupee of factory spend, has to keep borrowing or diluting to stay alive — and indeed borrowings rose nearly six-fold, from ₹155 crore to ₹913 crore (Inve data, 2026). That is the entire thesis the single ROE number hides.

Value — am I being asked to pay a sane price? A great business at a silly multiple is a poor investment. P/E, EV/EBITDA, and price-to-book are the standard tools; the discipline is to apply them within a sector, never across. A 40x P/E is cheap for a high-growth franchise and absurd for a cyclical commodity maker.

Delivery — does management do what it says? This is the filter the screeners forgot. It does not come from the balance sheet. It comes from comparing what management guided in past earnings calls against what actually showed up in later results — whether targets were hit, missed, or quietly dropped.

What does a worked screen look like?

Here is the same starting universe run through a loose screen and a working one. The counts are clearly-labelled illustrations, sized to show the shape of the funnel — not real counts from any one run.

The loose screen ends where it began. The working screen ends with roughly two dozen names — small enough to open each one's earnings calls and form a real view. Notice the OCF/EBITDA condition in the quality row: that one clause is what a Kaynes-shaped name fails. A business consuming cash while it books profit never clears a "cash converts to at least 70% of EBITDA" gate, no matter how good its ROE looks. The delivery filter then does the heaviest cutting at the very end, precisely because it removes companies that look good on paper but have a track record of talking better than they execute.

The order is deliberate. Quality first (cheap to apply, removes the most names), value second (keeps you from overpaying for the survivors), delivery last (the most labour-intensive, so you only do it on a list that has already survived two cuts).

How do you screen for things the screener can't see?

This is where almost every screening guide stops — at the ratios — and where the real edge begins.

Two companies in the same sector can post identical ROCE, identical growth, identical valuation. The screener cannot tell them apart. The difference is whether one management consistently delivers on guidance and the other consistently slips and then goes quiet. That gap only appears when you read the earnings-call record across several quarters.

Go back to Kaynes for the cleanest illustration, because the cash-flow problem and the delivery problem are the same problem. In its Q3 FY25 call management guided operating cash flow to turn positive by Q4 FY25. Q4 FY25 came: it was negative. By Q1 FY26 the line had become "significantly positive OCF" for FY26; by Q3 FY26 it was reaffirmed as positive "by year-end." FY26 closed — and consolidated operating cash flow was still negative; the target is now formally missed and re-guided to FY27 (Inve data, 2026). Four calls, the same destination, never reached. The working-capital target tells the same story — guided "sub-70 days," it was sitting at 139 days a few quarters later (Inve data, 2026). On the Q1 FY26 call management put it plainly: "by the time we reach the end of the year, I think we would probably perform on the expectations in terms of both OCF and as well as the working capital days, let's say, coming sub-70" (Kaynes Q1 FY26 concall). Sub-70 became 139. That is a said-versus-did gap no balance-sheet line will ever print for you, and it is exactly the kind of slip the delivery filter is built to catch.

Sometimes the dodge is subtler than a missed deadline — it is a quietly changed subject. On its Q4 FY25 call, IIFL Capital's management said its wealth-management book of "roughly close to about Rs. 800 crores - Rs. 900 crores of assets under management … should easily double in the next one year" (IIFLCAPS Q4 FY25 concall). By the next call the conversation had moved to a "Distribution AUM" of ₹35,700 crore; the call after that, to "Cross-sell assets" of ₹44,000 crore; and the original ₹800–900 crore bucket that was supposed to double was never mentioned again (Inve data, 2026). The target didn't fail loudly. It was swapped out for a bigger, different number while no one was keeping the old one's ledger. A screen sees none of this — the AUM line just keeps growing.

And the harder the commitment, the louder the silence when it's dropped. On its Q1 FY26 call, KRBL's management committed that an "independent reputed third party firm be appointed to conduct a thorough review … not later than 30 days" (KRBL Q1 FY26 concall) — a governance review with a 30-day clock on it. The next three calls came and went with no update on any of them — through Q4 FY26 the issue simply remained unaddressed (Inve data, 2026). A 30-day commitment that goes unmentioned for three straight quarters is not a forecast that came in light. It is a dropped commitment, and only the call record shows it.

This is genuinely hard to do by hand. Tracking guidance-versus-delivery for one company across eight quarters is an afternoon's work; doing it for a 25-name shortlist, every results season, is not something a person with a job sustains. It is exactly the kind of long-memory, cross-company pattern that has to be assembled, not screened. Inve's KPI Screener surfaces the operating KPIs management actually reports on calls — revenue growth, order book, capacity utilisation, NIM, working-capital days, and the rest — with their year-on-year and quarter-on-quarter trend and a data-confidence flag, so you can rank a sector on the operating numbers that lead the reported results. The delivery record itself — what was guided, what landed, what was quietly dropped — lives in the Promise Tracker.

The honest framing: the financial screen narrows 4,000 to 25. The delivery layer is what helps you choose among the 25 — and, just as often, what tells you which two of the 25 to throw out before you waste a weekend on them.

What thresholds should I actually use?

Resist the urge to copy someone else's numbers. The right thresholds depend on the sector, and the temptation to over-specify is itself a failure mode — twelve precise filters feel rigorous and usually just encode your biases.

A workable starting frame for a non-financial company:

• ROCE above 15% sustained over a full cycle, not a single good year.
• Operating cash flow at least 60–70% of EBITDA across three years — profit that actually converts to cash. (This is the one Kaynes fails, and it is the most important.)
• Interest coverage above 3x — the business can service its debt with room to spare (a low ratio is one early sign of a debt-trap stock).
• Sector-relative valuation in the cheaper half of comparable peers, not an absolute P/E cutoff.

For banks and NBFCs, throw most of that out — leverage is the business model, and negative operating cash flow is normal (a growing loan book consumes cash by design, so the Kaynes red flag would mislabel every healthy lender). Screen on net interest margin (NIM), gross NPA, return on assets, capital adequacy, and the trend in early-stress assets instead.

That last one is where a lender screen earns its keep, because two housing-finance and NBFC names can look alike on headline NIM and diverge entirely on asset quality. Repco Home Finance carried Stage-2 assets — loans showing early stress but not yet bad — at 9.7% in Q1 FY26, with management openly setting "a target of achieving GNPA percent of 2.5% and maintaining stage 2 numbers between 7-8% by year end" (REPCOHOME Q1 FY26 concall); the headline gross NPA itself improved from 3.30% to 2.50% across those quarters (Inve data, 2026). Contrast the NBFC arm of Aditya Birla Capital, whose combined Stage-2-and-3 book fell steadily — 4.25%, 3.78%, 3.70%, 3.03% from Q3 FY25 to Q2 FY26 (Inve data, 2026). One lender is working a known early-stress overhang down toward a stated target; the other's stress book is simply shrinking quarter after quarter. A NIM-and-GNPA screen treats them as peers. The Stage-2 trend is what separates them — and it is the line a manufacturing ROCE screen, run on a lender, would never even look at.

Whatever you pick, keep the count of filters small. Three or four sharp filters that each remove real names beat ten filters that each shave off a handful and together encode a portrait of a company that doesn't exist.

Where this approach could be wrong

The strongest case against leaning on the delivery layer is that it is a young record. Inve's transcript history runs roughly two years deep, so a guidance "slip" can be the honest shape of a capital-heavy company mid-build, not a character flaw. Kaynes is exactly that ambiguity: a manufacturer pouring money into OSAT and PCB capacity will burn operating cash and carry heavy working capital for a few years — that is a coherent, even likely, explanation, and the depressed cash flow could simply be the bottom of a J-curve. The delivery filter cannot tell you which it is. What it can do is force the question into the open: is this slip the normal physics of a build-out, or a management that habitually guides better than it executes? A short record is a strong prior, not a verdict — treat it as one. And the inverse failure is real too: a management can hit every soft target it sets precisely because it only ever guides things it knows it will clear. A clean delivery record on trivial commitments is its own kind of tell.

Frequently asked questions

A screen is not the analysis — it is the doorway to the analysis. Its only job is to take an impossible universe and hand you a list short enough to read, company by company, call by call. The filters most people use return the consensus list at consensus prices, then stop exactly where the interesting work starts: the forward record of whether management delivers. The Kaynes profit that never became cash, the IIFL Capital target that was quietly swapped for a bigger number, the KRBL 30-day commitment that three calls forgot — none of those show up in a ratio. They show up only when someone keeps the ledger.

So end on the owner's question, not the screener's: of the two dozen names that survive your three sieves, which managements have actually done, over the last several quarters, the things they told their owners they would do? See how the companies on your shortlist have delivered on past guidance — their call-by-call record is already in Inve's Promise Tracker.

Related reading

• Capacity Utilisation: Pre-Capex Signal for Stocks
• How to Analyse a Cement Company (India)
• How to Analyse a Pharma Stock in India