How the numbers are made

• Financial results — the machine-readable (XBRL) versions of the quarterly and annual results each company files with BSE. Annual history goes back to FY2019 where the filings are available.
• Prices — end-of-day prices. The history was loaded from NSE end-of-day files; new days are added from BSE's daily price file (the bhavcopy).
• Corporate actions — splits, bonuses, dividends, demergers and schemes, from BSE. Past prices are adjusted for splits and bonuses so a share's history is comparable; returns shown are price changes and do not include dividends.

The site covers the Nifty 200. Each company page names the latest results filing and the latest price date it uses. The full list of sources and their update schedule is on data sources.

Each filing is read line by line, and every figure stays tied to the accounting tag and the period it was reported for. Before anything is used, the filing's own arithmetic is checked: assets must equal liabilities plus equity, and income minus expenses must match the reported profit before tax.

The figures are stored exactly as filed and never edited. From them, the site builds one standard set — revenue, EBITDA, profit, debt, net worth and so on — using separate rules for ordinary companies, banks, NBFCs and insurers, because each files in a different format. When a filing reports a figure only indirectly, it is derived only if the result reconciles exactly with the filing's other lines; otherwise it is left blank rather than estimated.

Quarterly figures are the quarter column each filing reports. Where a filing reports only a year-to-date total, the quarter is the difference between two year-to-date totals; a quarter that cannot be worked out this way is left blank.

Most companies report two sets of results: consolidated (the whole group, including subsidiaries) and standalone (the parent company alone). Each page shows consolidated figures by default, and standalone only when the standalone history is longer. You can switch between the two on every company page.

Some years are filed only as consolidated results; those years are included in the consolidated history.

Each figure is compared with the company's own history and with the related figures in the same filing. A figure that jumps implausibly, or contradicts the others, is flagged at one of two levels:

• Suspicious — shown as filed, because it may be a real business event, but left out of the multi-year comparisons behind the written insights.
• Quarantined — judged to be wrong, and excluded from every chart, ratio and insight.

Earnings per share is checked against profit and the share count, and every correction records its reason. A company whose data has known problems can be removed from coverage altogether.

Compound annual growth is (last value ÷ first value) raised to 1 ÷ years, minus one. Three details make it honest:

• Actual years. The number of years is the real time between the two fiscal year-ends. A missing year never shortens the stated period, and after a change of year-end the period can be fractional — “over 2.3 years”.
• Positive endpoints. Growth runs between the first and the last years with a positive value. If the earliest year was a loss, growth is measured from the first profitable year, and the label names the years actually used.
• Context. When the period spans a merger, acquisition or demerger, the text says so. When one year's jump dominates the rate, or the starting year was unusually weak, the insight notes that the compounding is uneven or a recovery.

Transition periods. A company that changes its financial year-end files one longer year — for example 15 or 18 months. That period is shown in the full tables with its real span (“Oct 2024–Mar 2026, 18 months”) but is left out of growth rates, year-on-year comparisons and insights, because it is not comparable with a 12-month year.

Year on year compares a year with the year whose year-end is 11 to 13 months earlier. If there is no such year, the comparison is left blank rather than stretched across a gap.

Ratios are calculated from the standard figures, the same way for every company, rather than copied from the filing. Ratios that relate a year's result to a balance — return on equity (ROE), return on capital employed (ROCE), return on assets, asset turnover — use the average of the opening and closing balance where the previous year is available. Margins are a share of revenue.

Only the metrics that suit a business are calculated:

• Companies — margins, ROE, ROCE, debt to equity, interest cover, working-capital days, cash conversion, free cash flow and more.
• Banks — net interest margin, cost to income, credit cost, gross and net NPA, CASA ratio, credit-to-deposit ratio, and growth in advances and deposits. EBITDA and similar metrics are not shown for banks.
• NBFCs — margins, ROE, ROCE, debt to equity, and gross and net NPA where reported.
• Insurers — net premium growth, expense ratio, and benefits paid as a share of premium.

Peers are companies in the same industry that file in the same format — a bank is never compared with a manufacturer. If the narrowest group has fewer than two companies, the comparison widens to the industry, then the sector. A company's position is shown against the peer median, and rankings count only the peers that have the figure.

The short sentences on each company page are written by fixed rules — 63 for annual results and 15 for quarterly results — not by an AI model. Each rule compares one figure with the company's own history, its peers, or the previous year, and produces a sentence only when the comparison clearly holds. Every insight stores the numbers and the years it used, so it can be checked, and rebuilding from the same data produces exactly the same text.

The rules describe what happened; none recommends buying, selling or holding a share, and the wording is checked automatically for that.

Holds. When a comparison would mislead — for example, a year-on-year fall that is really a demerger moving part of the business out — that kind of insight is held back for the company until the rule can account for it. Each hold has a stated reason and a review date.

A company page is visible to everyone, but it is kept out of search engines until it has enough data to be useful: at least two years with revenue and profit, a latest year with revenue, profit and net worth, and at least half of the ratios that apply to the company. Pages are added automatically once they qualify.

Filings are sometimes restated, filed in unusual formats, or filed late; the checks above catch many problems but not all of them. Known gaps are listed on data sources. If a figure looks wrong, please report it.

For research and educational use only. Not investment advice. See the disclaimer.