Alternative Data in Indian Markets

Alternative data means any source beyond financial statements and price and volume. In the US the market for it is mature: satellite images of car parks, card transaction panels, web traffic, job postings, all sold by specialist vendors. In India the ecosystem is younger, and shaped by different data, different regulation and a different economy.

My view after years of looking at this: the most useful alternative data in India is not the expensive kind. It is the structured disclosure data that SEBI already requires companies to file and that NSE and BSE publish for free, which most investors never read. Here is what exists, what it costs, and how to test whether any source has an edge worth chasing.

1. IngestExchange filings, government releases.
2. MapTie every record to an ISIN and a filing timestamp.
3. Build the signalChanges, not levels: pledge delta, holding delta.
4. Back test after costsWalk forward, as of the filing time.
5. Paper tradeMonths of live data before any real capital.

A sensible alternative data pipeline. Step 4 kills most ideas, and step 2 is where most look ahead errors hide.

The Indian alt data landscape

The underused free signal: promoter pledging

Of everything available in India, promoter share pledging is the most underused relative to what it tells you. Promoters pledge shares to raise loans. A rising pledge says the promoter needs cash and cannot or will not sell equity. High pledges, often above 30% to 40% of the promoter's holding, have preceded several of India's best known corporate collapses, in housing finance, consumer goods and infrastructure.

Worked example: how a pledge turns into forced selling

A promoter owns 60% of a company whose shares trade at ₹500. They pledge half their stake, 30% of the company, against a loan at a 50% loan to value ratio. For every ₹100 of shares pledged, they borrow ₹50.

Illustration. Actual loan terms and margin triggers vary by lender.

The last row is the danger. Lender selling pushes the price down, which triggers more margin calls, which forces more selling. A 30% fall can become 60% for reasons that have nothing to do with the business. A screen that excludes companies with high or rising promoter pledges removes much of this tail risk, using data that costs nothing.

1. Day 0The event

   A promoter pledges shares, an insider buys, a bulk trade prints.

2. Same eveningBulk and block deals published

   Exchanges post the day's large trades after market hours.

3. Within 2 trading daysInsider trade disclosed

   For trades above ₹10 lakh in a calendar quarter.

4. Within 7 working daysPledge disclosed

   Creation or release of an encumbrance by promoters.

5. Within 21 days of quarter endShareholding pattern filed

   The full quarterly picture, including pledged shares.

Your signal can only start on the disclosure date. Back tests that use the event date are looking ahead.

Three free Indian alt data signals worth tracking

1. Shareholding pattern, quarterly. Every listed company files its shareholding pattern within 21 days of quarter end, in a SEBI prescribed format: promoter, foreign institutional, domestic institutional and public holdings, and shareholders above 1%. Two patterns are worth watching. Rising institutional ownership in a quality company with a stable promoter stake is a positive sign. A falling promoter stake together with a rising pledged share is the strongest distress signal in public Indian data.

2. Bulk and block deals, daily. A bulk deal is a client's trades in a stock totalling more than 0.5% of its listed shares in a day; exchanges publish them the same evening. Block deals are large negotiated trades in a separate window; since SEBI's circular of 8 October 2025 the minimum order is ₹25 crore, within 3% of the reference price, with same day disclosure. A buyer who keeps appearing in the same stock across quarters is telling you something about conviction.

3. Insider trades, event driven. Under SEBI's insider trading regulations, promoters, directors and designated employees must disclose trades worth more than ₹10 lakh in a calendar quarter within 2 trading days. Clusters of buying by several insiders in a short window have been studied as a positive signal. Insider selling is noisier; it can be tax planning or personal liquidity.

How to evaluate whether alt data has genuine edge

Every source should pass four tests before you build on it:

• Is there an economic reason? Why should this predict returns? Without a plausible mechanism, a good back test is probably noise.
• Was it available at decision time? Use the filing timestamp, not the event date. A pledge created on the 1st and disclosed on the 9th can only be traded from the 9th.
• Does the edge survive costs? Many alternative signals decay within days. Fast signals mean frequent trading, and STT, charges and impact eat short edges first.
• Will it persist? Easy to get signals get crowded fast. The edge in public data lasts only while it stays tedious to assemble.

Worked example: is the signal real?

The standard yardstick is the information coefficient: each month, the rank correlation between your signal and the next month's returns across stocks. Say a pledge change signal shows an average monthly IC of 0.03 with a standard deviation of 0.10 over 60 months. Its t statistic is 0.03 / (0.10 / √60) ≈ 2.3. Promising, not proven, and if you tried ten variants of the signal to find it, not even promising yet. Module 5.1 explains why.

What does not work as alt data in India yet

Two kinds of data that work in the US disappoint in India, because the underlying economy is different.

Card transaction panels. In the US, deep card databases feed consumer research. India runs on UPI and cash. Aggregate UPI volumes are published monthly, but merchant level detail is not available to outsiders, and card data alone misses most spending.

Satellite views of factories and shops. These work where companies operate from a few identifiable sites. Indian manufacturing is often spread across contract partners and shared facilities, so the signal for a typical mid cap is weak. Satellite works for ports, airports, mines and a handful of large integrated plants, not for the broad market.

Common mistake: paying for novelty

The common mistake is buying data because it sounds advanced. An expensive feed with a beautiful dashboard is still worthless if its information coefficient is zero after costs. Ask any vendor for an out of sample test on dates you choose, with the data as it was available on each day. Most cannot provide it.

The second mistake is the opposite: ignoring free data because it looks boring. The quarterly shareholding file and the daily pledge disclosures will never appear in a pitch deck. They have explained more Indian blow ups than any satellite.

How this shows up in RupeeCase

Questions readers ask