🚨 Red Flags Before They Become Headlines

Annual Reports Are Designed to Impress, Not Reveal

Every listed Indian company publishes an Annual Report after the end of each financial year. Every investor knows they should read it. Almost no one does — not because they don't want to, but because a typical Annual Report runs 200 to 400 pages of dense text, financial tables, legal disclosures, and corporate photography of smiling executives shaking hands at ribbon-cutting ceremonies.

The manageable parts — the Chairman's Letter, the MD&A, the business review — are written by communications teams whose job is to frame the year's results in the most favourable light possible. A year of declining margins becomes a story about “strategic investment for future growth.” A year of cash burn becomes “aggressive capex in high-return projects.” A related party transaction with a promoter-owned entity appears in a footnote in the Notes to Accounts, expressed in crores, without context about market rates or the nature of the relationship.

Consider what happens when an auditor raises a qualification. In the narrative section of the Annual Report, there may be no mention of it at all. You would need to read the Auditor's Report — which appears on page 180 of a 300-page document, written in the formal language of accounting standards — to find it. Then you need to understand what the qualification means in plain English, whether it is material, and what the investment implication is.

Or consider related party transactions. The Notes to Accounts list every material RPT: which entity, what relationship, what nature of transaction, what amount. But they don't tell you whether the rent paid to the promoter's family trust is at market rates, or whether the management fee paid to the holding company is a reasonable cost or a wealth transfer mechanism. That assessment requires judgment, and individual investors rarely have the time or training to apply it systematically across all material RPTs.

Red Flag Detector applies that judgment systematically, across all seven categories of concern, in one structured scan.

Red Flag Detector: Adversarial Forensics with a Scorecard

Red Flag Detector is adversarial by design. The AI is explicitly instructed not to balance its findings — it is not asked to say “while the company has related party concerns, its earnings quality is strong.” It is asked to find what is wrong, quantify it, and report it without softening. The investor can weigh positives against negatives. The template's job is to make sure the negatives don't get buried.

The output is structured across 7 sections and closes with the N/18 Red Flag Scorecard and an overall rating:

The N/18 model assigns up to 3 flags per category across 6 substantive categories (18 total). This normalisation allows cross-company comparison and makes the composition of the score as important as the total — a company with 4 flags all in the Auditor section is more concerning than one with 4 flags spread across four categories.

What AIA Pre-fills Automatically

Before the AI reads the Annual Report, Finmagine AI Advisor has already extracted 5 years of financial data from Screener.in and injected it into the prompt. This pre-filled data directly drives three of the seven sections:

• Section 2 (Balance Sheet Stress Test): 3-year receivable days, inventory days trends from Screener financial tables
• Section 3 (Cash Flow Integrity): 3-year PAT vs CFO comparison directly from the Cash Flow table
• Section 6 (Governance): Promoter shareholding trend from the Screener shareholding pattern table, FII/DII trends

Additionally, AIA injects the Annual Report links from Screener.in's Documents section. Claude reads the most recent link automatically. Sections 4 (Related Party Audit) and 5 (Auditor Signals) require the Annual Report and cannot be completed from Screener data alone.

What Red Flag Detector Checks: Section by Section

Section 1 — Earnings Quality Scan

This section examines whether the reported profit number is a reliable measure of business performance, or whether it is being propped up by accounting choices that are technically permitted but economically misleading. Five specific checks from the Annual Report's Notes to Accounts and Auditor's Report:

Section 2 — Balance Sheet Stress Test

A 3-year structured table comparing balance sheet health metrics against Screener's financial data. This section catches the slow deterioration that doesn't show up in the headline P&L:

Goodwill, Capital Advances, Loans to Subsidiaries, and Contingent Liabilities are all from the Notes to Accounts — not available in Screener's financial tables. This section requires Annual Report access to complete fully.

Section 3 — Cash Flow Integrity

The single most important financial check in the entire scan. This section uses AIA's pre-filled Cash Flow table to build a 3-year comparison of PAT vs CFO (cash from operations):

Flag thresholds: CFO/PAT below 0.7 for 2+ consecutive years; FCF consistently negative while PAT grows; large working capital drag absorbing more than 30% of EBITDA; capex funded by debt while simultaneously paying dividends.

The bottom-line question this section must answer: Is PAT growth being supported by real cash generation, or are earnings accrual-driven? Section 3 can be completed from pre-filled Screener data without Annual Report access.

Section 4 — Related Party Audit

This section lists every material Related Party Transaction (RPT) above ₹10 Cr or 1% of revenue from the Annual Report's Related Party Disclosures. For each transaction:

• Entity name and relationship (subsidiary / associate / promoter group)
• Nature of transaction (rent, loan, sales, management fee, royalty, brand fee)
• Amount in ₹ Cr
• Flag: CLEAN or CONCERNING, with a one-sentence reason

The specific patterns the AI is trained to identify as CONCERNING:

• Rent paid to promoter-owned entities — wealth transfer from the listed company to the promoter family without shareholder approval at market rates
• Loans to group companies at below-market rates — the listed company is effectively subsidising related entities
• Management fees or brand royalties flowing to the holding company — a recurring extraction mechanism that compounds across years
• Sales to related parties at above-market prices — inflating revenue by selling to captive buyers who have no choice
• Purchases from related parties at above-market prices — compressing margins by buying from related entities at inflated costs

Section 5 — Auditor Signals

The auditor is the gatekeeper who is legally required to flag material concerns. When something appears in the Auditor's Report, it passed a professional judgment threshold. When something is absent from the Auditor's Report despite being visible in the financial data, that too is a signal. Six checks:

Section 5 is completely dependent on Annual Report access. When BSE 403 blocks the link, all five checks are marked “UNKNOWN — data gap” and the AI explicitly marks this as a severe data gap in its output.

Section 6 — Governance & Promoter Behaviour

This section combines AIA's pre-filled shareholding data with Annual Report corporate governance disclosures. Six checks:

The promoter holding trend (from Screener shareholding pattern) is particularly useful here: steady reduction in promoter holding over 3–5 years, even if modest in any single year, signals systematic distribution that management never needs to disclose explicitly.

Section 7 — Red Flag Scorecard

The final section tallies all flags into the N/18 scorecard and delivers the overall rating, most critical finding, and one sharp question for the next earnings call:

How to Read the N/18 Score — Including the Data Gap Nuance

The N/18 score is more useful than a simple good/bad rating because it tells you where the problems are concentrated, not just how many there are. But it requires careful reading in one specific situation: when the Annual Report was inaccessible.

The Data Gap Problem

Annual Report links on BSE sometimes return 403 Forbidden errors because they require a browser session that the AI doesn't have. When this happens, Sections 4 (Related Party Audit) and 5 (Auditor Signals) are entirely blocked. The AI marks all items in those sections as “UNKNOWN — data gap” and counts them as flags, because an unverified area of concern is still a concern.

The AI is instructed to explicitly distinguish data-gap flags from confirmed findings in its output. A well-generated Red Flag Detector result will say something like: “TOTAL FLAGS: 7/18 (4 confirmed, 3 data gaps — BSE 403 blocked Sections 4 and 5)”.

How to Handle a BSE 403

If your result shows significant data gaps in Sections 4 and 5, you have two options:

1. Download and upload the Annual Report PDF directly. Find the Annual Report on the company's investor relations page or on BSE/NSE, download it, and attach it to Claude before submitting the prompt. This is the most reliable approach for thorough due diligence.
2. Use the fallback link. AIA injects multiple Annual Report links (typically 10–15 years of reports from Screener). If the most recent link is blocked, Claude will try the second-most-recent. For some companies this works — the previous year's Annual Report is accessible even when the current one isn't. The score will be based on the prior year's report, which is still useful context.

The Composition Matters as Much as the Total

Two companies with a score of 4/18 can have very different risk profiles depending on where those flags fall:

• 4/18 with all flags in Earnings Quality: The company is probably managing earnings — watch the next 2–3 quarters carefully, but the governance structure may be intact.
• 4/18 with 2 flags in Related Parties and 2 in Auditor: Much more serious. Auditor concerns and RPT flags together signal potential governance issues that go beyond accounting presentation.
• 4/18 with flags spread across Cash Flow, Balance Sheet, Governance, and Auditor: Systemic concern — no single category is catastrophic but the breadth suggests structural issues rather than one-time events.

Acting on Your Red Flag Detector Result

🌺 CLEAN (0–2 flags)

No material concerns found in the Annual Report. This doesn't mean the stock is a good investment — it means the governance and earnings quality signals are clean. You can focus your remaining due diligence on valuation and competitive dynamics without worrying that you're buying into a governance problem.

Run annually. A CLEAN company can develop flags in a subsequent year. Red Flag Detector is designed for the cadence of Annual Report publication — once a year, every year, for your core holdings.

🌡 WATCH (3–5 flags)

One or more specific areas of concern, but not across the board. The output will identify exactly which sections have the flags. Your next step is to investigate those specific areas:

• If the flags are in Earnings Quality: watch the next two quarters for margin normalisation. A one-year margin spike is less concerning than a trend.
• If flags are in Related Parties: dig into the specific RPTs named. Are they arms-length? Has the amount been growing? Check management's explanation in the AGM or concalls.
• If flags are in Auditor: read the full Auditor's Report, including the KAMs. Understand what the auditor was focused on and whether the concern has been resolved.

🔴 CAUTION (6+ flags)

Significant concerns across multiple sections. This verdict does not mean the stock will fall — markets can sustain high-flagging companies for extended periods if the narrative is strong. It means the due-diligence burden is very high, and the risk/reward at the current price may not compensate for the identified concerns.

Two specific escalation actions:

1. Run Forensic Governance. If the flags are concentrated in Sections 4 (RPTs) or 5 (Auditor), escalate to Forensic Governance, which runs a deeper investigation using concall transcripts alongside Annual Reports. Forensic Governance can find whether management has explicitly addressed the RPT concern in analyst Q&As, and whether evasion patterns are visible.
2. Download and upload the full Annual Report PDF. If data gaps are inflating the score, eliminate that uncertainty before making any decision. A 🔴 CAUTION based on 8 data gaps is completely different from 🔴 CAUTION based on 8 confirmed findings.

Red Flag Detector vs Forensic Governance vs IPO Decoder

All three are forensic templates. The right choice depends on what the company is and what question you're answering.

The recommended workflow: Red Flag Detector first (5–10 minutes, N/18 scorecard) → if Sections 4 or 5 have confirmed flags, escalate to Forensic Governance (20–40 minutes, concall evidence added) → for pre-listing companies, start with IPO Decoder instead.

Run Red Flag Detector annually on every company you hold with a significant position. Annual Reports are published once a year and the template is specifically designed for that cadence. A company that scored CLEAN last year can develop material flags in a single Annual Report year — and you want to know before the market prices it in.