How FinnGuru works

1. Evidence first

For each stock FinnGuru collects prices and returns, quarterly and annual results, results dates and earnings estimates, exchange filings, news headlines, shareholding and industry peers. Code turns these into numbered facts, each with its source and date. The AI can only use these facts.

2. Specialist AI agents

News agent

Reads recent headlines and exchange filings, merges them into distinct stories, and labels each as fact, opinion or speculation.

Results agent

Reads quarterly and annual numbers and writes a short results note: growth, margins, cash flow, debt, and one-off items that distort comparisons.

Price-move agent

Explains 1-month, 3-month and 1-year moves versus NIFTY 50 and the industry, and links big single-day moves to dated events, or says the cause is unclear.

Brief agent

Pulls it together: what’s happening, why, what changed, where the evidence is strong or weak, and what to watch next.

Agents run on the best LLMs for each task. If one agent fails, the others still run and the page says what’s missing. Hindi versions are translated from the checked English brief.

3. Every claim is checked

• ✓Every claim must cite at least one numbered fact that exists
• ✓Every number in a claim must appear in the facts it cites, with the same sign
• ✓Quarter labels (like Q1 FY27) must match the cited facts
• ✓Buy/sell/hold language, price targets and words like “cheap” or “undervalued” are removed
• ✓Accounting distortions (such as large exceptional gains) are detected in code and flagged to the agents

Claims that fail a check are removed, not rewritten. In the app, tap the small numbers after any sentence to see its source.

Chart intelligence (Trader)

Move Map

Big single-day moves (unusually large for that stock) matched to results, dividends, filings and headlines dated around them. One-session jumps over 25% with no results nearby are flagged as likely corporate actions, like demergers or splits.

Results Reaction

Price change from the close before each results announcement to the close of the second session, compared with NIFTY 50, plus the volume spike and 5-session drift.

Delivery Pulse

NSE delivery % versus the stock’s 20- and 60-session averages and its industry. A spike means delivery volume at least 2× its average and delivery % at least 10 points above average. Days when delivery was high across the Nifty 500 are marked market-wide.

Crowded Price Zones

Volume-at-price estimated from daily high–low ranges: the most-traded zone and the price range holding about 70% of volume.

These describe what happened. They aren’t buy or sell signals, support/resistance calls or predictions.

FinnGuru Signals

Five scores from 0 to 100, each a percentile against the Nifty 500. Higher is more favourable on that factor. They describe where a stock stands today; they’re never combined into one rating and aren’t predictions.

Quality

Return on equity, operating margin, net profit margin, debt-to-equity (lower debt ranks higher).

Valuation

P/E and P/B versus industry peers. A lower multiple ranks higher. Loss-making companies get no P/E rank.

Growth

Revenue and earnings growth in the latest quarter versus a year earlier.

Momentum

3-month and 1-year returns, and price versus the 200-day average.

Stability

1-year volatility and the largest drop from a peak. Steadier ranks higher.

What FinnGuru is not

FinnGuru is not a broker, a tips service or a SEBI-registered investment adviser or research analyst. It doesn’t give buy, sell or hold recommendations or price targets. Read the full disclaimer.