How a number gets here

From the filing to your screen, checked on the way.

  1. Fetched from NSE
  2. Read as filed
  3. Checked against its own total
  4. 48,036 filings, the same test
  5. Served with its source
  • 36,566 add up
  • 11,455 one block rebuilt, flagged
  • 12 total rebuilt, flagged
  • 3 don’t add up, kept as filed

1,471,548 figures. 0 differ.

Every SEBI portfolio-manager report since Jan 2018, read back.

Rounding, shown

Filed 0.1011

Both round to 10.11%.

Ours 10.1062

What we leave out

Only what was filed, by whom and when.

  • Share prices
  • Market values
  • Scores
  • Ranks
  • Ratings
  • Forecasts
  • Recommendations

Every build is kept

Cite it today, fetch it again later: version=2026-10-08.

version=2026-10-08

Found a wrong number?

  1. Send the link
  2. We check the filing
  3. Fixed in the next build, dated

hello@brihas.in

In detail

The checks, counted

We fetch each filing from the exchange or regulator that publishes it, and keep a copy.

We read numbers and labels as tagged, and show the filed label beside any we rename.

A shareholding pattern has three blocks: the promoter group, the public, and non-promoter non-public holders. Together they must equal the filed total.

How the 48,036 filings in the build of 8 Oct 2026 came out:

Shareholding-pattern filings in the build of 8 Oct 2026, by check result
FilingsResult
36,566All three blocks (promoter group, public, non-promoter non-public) are tagged and add up to the filed total.
11,455One block is not tagged. In 11,198 of these it is the non-promoter non-public block, which is usually empty. We take it as the filed total minus the other two blocks, and flag the row.
12The filed total is missing or unusable. We use the sum of the three filed blocks, and flag the row.
3The blocks do not add up to the filed total. We keep the filing as filed, and flag it.
48,036All filings, from 3,015 companies under 3,280 ISINs (46,896 from NSE, 1,140 from BSE)
One filing, line by line

A total that adds up can still hide a wrong line, so we also compare rows with the filing.

Reliance Industries, quarter ended 30 Jun 2026, filed 16 Jul 2026, against our database row:

Our database row against the filing, Reliance Industries, quarter ended 30 Jun 2026
LineOur databaseThe filingResult
Promoter and promoter group6,715,496,0966,715,496,096Same
Public6,587,575,7586,587,575,758Same
Mutual funds1,344,430,2201,344,430,220Same
Insurance companies1,224,269,4211,224,269,421Same
Resident individuals up to Rs. 2 lakh952,958,091952,958,091Same
Foreign portfolio investors (I + II)2,270,149,1852,270,149,185Same
Total shares13,303,071,85413,303,071,854Same

Open the filing

Filings tag percentages as four-decimal fractions; we compute ours from share counts.

Rounded lines can add to 99.99 or 100.01. Share counts always add exactly, so our tables are ruled on share counts.

Read back against SEBI

Portfolio managers’ monthly reports are published by SEBI. We parsed every report from Jan 2018 to Aug 2026.

We then read each figure back against SEBI’s records: in the discretionary-assets table, 1,471,548 were compared and 0 differ (checked 23 Sep 2026).

This shows we reproduce SEBI’s records exactly. It does not show that SEBI’s figures are right.

Where SEBI’s own publications disagree with each other, we keep both as filed and flag them.

These reports are not served yet. We will serve them once SEBI agrees.

IPO anchor books

Each anchor book in the build of 8 Oct 2026 passed these checks. The number is how many failed:

  • 0 lock-in tranches that are not half of the anchor book.
  • 0 issues where the anchor shares do not add up to 100% of the anchor portion.
  • 0 duplicate allocation lines.
  • 0 issues with a gap in the allocation list's serial numbers.

Lock-in end dates follow from the listing date and the lock-in rule. They are estimates, and we label them so.

Dates, versions and speed

Every build is dated and kept; this site shows the build of 8 Oct 2026.

The API will answer version=YYYY-MM-DD with the numbers it gave on that day, so a figure you cite can be fetched again.

API, MCP and CSV carry the same value and the same source object.

Speed is measured, not claimed. One row takes 0.35 ms to query on our build machine, median of 40 runs.

Network times will be published on a status page once the API has run for seven days.

A free sample pack

For funds: name your companies, get a workbook built from their filings.