Philip A. Fisher's framework — Quality × Price × Discipline — applied to 65 watchlist names.
UNIVCABLES_ADDENDUM.md, in the format of the Atlanta addendum. Excluded at G4: the company has never held an earnings call, and the main run excluded all four no-call names on this gate. Stated, not hidden: the frozen wording also accepts “equivalent written disclosure of an adverse event”, which the main run did not test for those four; Universal arguably meets it through its Feb-26 and Aug-26 capex-delay filings. Kept as a FAIL for consistency and flagged. The verdict does not depend on it. Gates G1, G2, G5, G6 pass; G3 is triggered, unresolved (two-way trade with the associate Vindhya Telelinks ≈6.6% of revenue, the Atlanta pattern). D3.1 and D3.2 score PARTIAL, but Dimension 3 has 3 FAILs (D3.5 margin vs peers, D3.6 profitability leadership, D3.8 no recurring revenue), Dimension 4 has 3 FAILs (D4.1, D4.4 false grouping with 20–33% ROCE peers, D4.7 leverage — debt-service cover 0.70), and the total is 10 FAILs against the 6-FAIL line. IE 5 of 33.Philip Fisher was an American investor (1907–2004) whose method was qualitative before it was numerical: judge the business and the people first, the price second, and hold only while you keep checking. His framework has three layers — Quality × Price × Discipline — and this screen turns them into 6 pass/fail gates and 33 scored criteria, applied to all 65 names.
A gate is a veto. Fail one and the name is out, no matter how good everything else looks. These are cheap to check, so they run first.
| G1 | Operating history — at least 3 years actually trading and 1 full year of operating profit. Excluded nobody. |
| G2 | Circle of competence — can this business be judged from documents? Fisher refused property development and anything driven by consumer fashion, because he was repeatedly wrong there. Excluded KOLTE. |
| G3 | Integrity veto — any serious question over trusteeship: a SEBI action, an auditor resignation, promoter shares pledged above 10%, an undisclosed related-party deal. Excluded 7, including CARE and NRBBEARING. |
| G4 | Disclosure symmetry — can you see how management talks in a bad quarter? Needs two consecutive periods covering an adverse event. No transcript record means you cannot know, and Fisher's rule is that unknowable means drop it. Excluded 12 — the single most destructive gate. |
| G5 | Employee treatment — no adverse labour finding. Found nothing, which is not the same as verifying good treatment. |
| G6 | Evidence sufficiency — at least two independent sources on management quality. Company self-report counts as one, however many transcripts it fills. Not applied — see the limitations panel. |
What this looked like in practice: ASTEC ran an active investor-call programme through its profitable years and has published nothing in the 46 months since its losses began. HALDYN, MONARCH, AMIC and NITTA hold no calls at all. PARAS holds calls but publishes no transcripts.
Fisher held that two questions disqualify a company on their own, however cheap the stock is.
| D3.1 | The summary question — "What can this company do that others could not do about as well?" A specific, evidenced answer is required. "Scale", "quality" or "relationships" with no mechanism is a fail. |
| D3.2 | Entry barriers — can newcomers start up and displace the leaders? Free entry is a fail. |
Several names answered against themselves. WHEELS claims a "conversion model" as its moat while its own risk section says the business is "raw-material pass-through, not conversion." DYCL's profile says "no brand, no pricing power." DCMSHRIRAM's says "the businesses do not reinforce each other" — which is a negative answer to D3.1 by construction.
Seven criteria on whether the business is efficient and self-funding. This is where most of the universe died, and it died on cash rather than on price.
| D1.1 | Cost position — operating margin at or above the median of named competitors, over 5+ years. |
| D1.2 | Shakeout test — did it stay profitable in the worst year of the cycle? Tests survival in a downturn, not whether a loss ever appeared. |
| D1.3 | Internally funded growth — three-year cash from operations covers capex plus working capital. Repeatedly issuing equity to fund operations is a fail. |
| D1.4 | Margin defence — margin held by cost engineering or mix (pass), or by price rises, duties and commodity luck (partial at best). |
| D1.5 | Technical effort — measured by output (revenue from products launched in the last 5 years), never by R&D spend. |
| D1.6 | Financial control — companies with real control do not surprise their shareholders. At most one unflagged negative surprise or restatement in 12 quarters. |
| D1.7 | Capital conservation — working-capital days stable or improving. Debtor days rising more than 30% is a fail. |
The pattern: AZAD consumed ₹1,009cr of cash over three years while converting only 7% of operating profit into cash. CONCORDCS let debtor days go from 67 to 210. GARWARE's working-capital days went from 7 to 146 in a single year. None of these is visible in a P/E.
| D2 | People (8 criteria) — the salary gradient between the top executive and the next three; whether the company depends on one person; succession from within; D2.4 the delivery scorecard, guidance promised against actual delivered; whether management admits specific mistakes; and whether it can name its most important internal long-range problem. |
| D3 | Business characteristics (10) — which moat route it uses, margin versus the closest competitor, whether it leads on profitability rather than size, recurring revenue, and D3.9 growth duration: how many years of runway are actually visible. Under 3 years is a fail. |
| D4 | Price (8) — not "is it cheap" but is there a specific belief the market holds that the evidence contradicts? If you cannot name one, the stock can never rate better than rung 2. |
Note what is deliberately excluded from D4: the 52-week range, the distance from a peak, and the size of a drawdown are not inputs in either direction. Fisher: "if the fundamentals are bad enough, it may still be very high at 24."
Quality and price are crossed on a five-rung ladder. Quality is "HIGH" only if the company has at most ONE fail across all 25 quality criteria.
| Rung | Quality | What the market is paying | Action |
|---|---|---|---|
| 1 | HIGH | Below what it merits — a specific gap you can name | Buy |
| 2 | HIGH | Roughly in line | Buy |
| 3 | HIGH | Above merit | Hold only — no new money |
| 4 | Not high | At or below merit | Excluded |
| 5 | Any | Far above merit | Excluded — "most dangerous group of all" |
This is why two fails is fatal. Two fails means quality is not HIGH; not-HIGH can only land on rung 4 or 5; and both are excluded. A cheap price cannot rescue it, and that is the entire point of the ladder. Tier A / B / C on the survivor cards is only about position sizing — A is the most financially robust (cap 20%), B established (8–10%), C small or one-man (5%).
GATE_PASS.md, D3_MANDATORY_PASS.md and PORTFOLIO_8.md.Three survivors. The criteria pre-committed to this outcome: "fewer than 8 is allowed and expected." Fisher's own funnel ran ~250 considered to 1 bought.
The screen's sharpest finding is not the survivor list. It is that most of this universe fails on cash conversion, not multiples. The 28 names with a published three-year cash-flow series, of the 29 that reached financial scoring — Yash Highvoltage has none, and was scored IE on D1.3. Bar colour is the final verdict, not the cash figure — several names with strong cash were excluded for other reasons.
Source: Screener consolidated cash-flow statements, FY24–FY26. Green = survivor, amber = near-miss, red = excluded. CARE is absent: the 2-Sep correction moved it to a gate exclusion, so it never reached this stage.
Un-blinded only after all scoring was complete, exactly as the blinding protocol requires.
DEEP is Deep Polymers, not Deep Industries — ten years of the wrong company's P&L were parsed before the name-echo check caught it.min(PAT)<0 flagged its FY25 loss — but revenue GREW 35% that year, so it was a write-off during growth, never a cyclical trough. D1.2 asks whether a company survives a downturn. Re-checked against the other D1.2 exclusion: GPIL's is correct — its loss years carried revenue of −17% and −9%, a real steel trough. DEEP stays excluded on D1.6 + D2.4 (two Q4 write-offs, the second after an explicit "nothing else left" assurance), so the verdict is unchanged and the reasoning is now right.Sources: SCREEN_CRITERIA.md (frozen 15-Aug-2026) · GATE_PASS.md · D3_MANDATORY_PASS.md · PORTFOLIO_8.md · PHILOSOPHY.md and the 23-file chapter chain in 01_chapters/.
Financials pulled from Screener consolidated pages, 2-Sep-2026, each verified by company-name echo. Standalone artefact — not wired to the main dashboard or the deploy repo.