The Filing Is Late. The Footprints Arrive Every Day.

Alternative data is most useful when it tells you what to investigate before the next filing—not when it pretends to predict tomorrow’s share price.

A quarterly result is rich in accounting detail, but it arrives after the business activity occurred. Between two results, customers register vehicles, move goods, pay taxes, open demat accounts, download apps, visit stores and respond to job openings. Those footprints can help an investor test whether the operating story is strengthening, weakening or simply changing shape.

They can also create false confidence.

A rise in vehicle registrations is not automatically a rise in an automaker’s revenue. More e-way bills do not prove that a particular logistics company gained volume. App installs do not reveal customer quality. Hiring can mean expansion, replacement or an expensive project that never earns its cost of capital.

The edge does not come from possessing an unusual chart. It comes from knowing exactly what the chart measures, when the observation became available and how it connects to the economics of the company.

What counts as alternative data?

For public-equity research, conventional data usually means:

• financial statements and notes
• exchange filings
• management commentary
• shareholding disclosures
• market prices and volumes
• consensus estimates

Alternative data is evidence outside that standard packet. It may be official and highly structured, such as GST statistics. It may be a live public counter, such as a company’s job openings. It may describe the whole economy, an industry, a company or merely public attention.

The word “alternative” does not mean informal or unreliable. VAHAN is a government registration system. CDSL publishes depository statistics. AMFI publishes mutual-fund data. IRDAI publishes insurance activity. What makes the data alternative is the research question it answers, not whether the source is official.

The Altys India alternative-data map

As of 4 September 2026, the Altys warehouse contained:

Warehouse measure	Current snapshot
Enabled source families	20
Defined series	56
Series with observations	51
Point-in-time observations	420,694
NSE symbols mapped through the entity registry	126
Earliest observation date	March 2012

The coverage is intentionally uneven because the underlying sources are uneven. A monthly government workbook can provide a long archive. A live careers page may only become historical from the day someone begins collecting it.

The current data families include:

Research question	Example Altys evidence	What it does not prove
Is mobility demand changing?	VAHAN national and listed-maker registrations; FADA category retail	Factory dispatches, revenue or profit
Is formal goods activity changing?	Gross, net and domestic GST; e-way bill count and assessable value	One company’s growth
Is transport demand changing?	Petrol, diesel, ATF and LPG consumption	A listed operator’s market share or margin
Is market participation changing?	CDSL beneficiary accounts; AMFI equity net flows	Unique investors or future returns
Is insurance activity changing?	Life first-year premium; non-life gross direct premium	Value of new business or underwriting profit
Is a company expanding capacity or teams?	Live careers postings and store counts	Productive growth or incremental ROCE
Is digital adoption or attention changing?	App installs and ratings, web rank, Wikipedia and public-channel activity	Paying customers, retention or unit economics
Is a specific risk emerging?	Rating-action breadth, GRAP restrictions, US drug recalls	A final governance or investment verdict

The Altys alternative-data hub links the public tracker pages and explains the source-specific caveats.

The hierarchy investors often skip

Every signal belongs at one of three levels.

1. Economy-level data

GST, e-way bills, fuel consumption and demat accounts describe broad activity. They can establish the background against which companies operate.

If gross GST grows 8%, that does not mean every consumer company grew 8%. Inflation, imports, compliance, formalisation and sector mix all affect the national figure. It is a prior, not a company result.

2. Industry-level data

FADA vehicle categories and insurance-industry totals narrow the question. Passenger vehicles can behave differently from two-wheelers. Standalone health insurers can grow differently from the non-life market.

Industry growth still does not identify the winner. A company can lose share in an expanding market or gain share with discounts that damage economics.

3. Company-linked data

Maker registrations, app activity, jobs, stores and company-specific recalls sit closer to a listed business. But mapping is work.

One listed parent can have several operating entities and brands. One registration label may change after an acquisition. An insurer may be an unlisted subsidiary of a listed parent. A company’s app can be relevant to engagement while remaining only one part of the revenue model.

“Company-linked” is therefore not the same as “company revenue.”

Why point-in-time history changes the answer

Imagine that a public dashboard currently shows 100,000 registrations for July. Late records arrive and the dashboard revises July to 104,000 in September.

Both values can be legitimate:

• 100,000 was the figure an investor observed in early August
• 104,000 is the better current estimate of what happened in July

A conventional historical download often preserves only 104,000. A backtest can then behave as though the revised value was available before it existed. The result looks more accurate because it quietly borrowed information from the future.

Altys stores two clocks:

Observation date: the period the value describes
Availability date: when that value or vintage became knowable

For rolling public files, Altys also retains the content hash and source snapshot. For current-state sources—such as live job postings—daily collection creates a history that the source itself does not provide.

This is why alternative-data infrastructure matters more than a one-time scrape. The scarce asset is the dated sequence.

Read why point-in-time data matters for the same problem in financial statements and backtests.

Five examples of a responsible research bridge

Vehicle registrations → auto economics

The India auto sales tracker recorded 2.43 million national VAHAN registrations in August 2026, up 17.4% from the August 2025 value in the current snapshot.

That is evidence of stronger registration activity. The next questions are category mix, manufacturer mapping, market share, discounts, dealer inventory, exports, realization and margin. Only then can the signal meet an earnings model.

GST and e-way bills → demand context

Gross GST collections for July 2026 were ₹2.11 lakh crore, 7.9% above July 2025 in the current series. E-way bill count for the same month was 137.9 million, up 6.0% year on year.

Together, the series suggest expanding nominal tax and goods-movement activity. They do not say whether a portfolio company gained volume or earned an attractive incremental margin.

That gap is the research opportunity.

Demat accounts and fund flows → participation

CDSL reported 188.0 million individual beneficiary-owner accounts for July 2026, 16.8% above the year-earlier count in the Altys series. AMFI’s open-ended equity-scheme net inflow was ₹24,697 crore for the month.

The first is an account stock. The second is a monthly flow. One person can own multiple accounts, and CDSL is not the entire depository system. Neither measure is the same as DII cash-market purchases.

Still, the pair can help research exchanges, depositories, brokers, asset managers and registrars—after the analyst translates participation into each company’s revenue and cost model.

Insurance premium → growth quality

July 2026 life first-year premium was ₹47,005 crore, 20.7% above July 2025 in the current Altys history. Non-life gross direct premium was ₹31,398 crore, up 5.7%.

These are operating activity measures with different definitions. Life value creation needs product mix, persistency, margins and capital. Non-life value creation needs pricing, claims and reserve adequacy. Premium can grow while economics deteriorate.

Hiring and apps → forward-only evidence

Live openings vanish when filled. App stores do not offer a clean historical endpoint for every useful counter. If an investor begins collecting only after a company surprises the market, the prior path is gone.

That makes forward-only data valuable and dangerous. Valuable because daily vintages create proprietary history. Dangerous because the history begins when collection begins, not when the company began operating.

A chart must disclose that inception boundary.

The seven-question alternative-data checklist

Before a signal enters a company note or model, ask:

1. What exactly is counted? Registrations, dispatches, accounts, people, documents, rupees, postings or page views?
2. Which entity and geography are covered? India, one state, one platform, one subsidiary or the whole listed group?
3. When did the value become available? Do not replace release time with period end.
4. Can the source revise history? If yes, preserve vintages and decide which one a backtest may use.
5. What is the denominator? A percentage change without market size, coverage and seasonality can mislead.
6. What connects the signal to profit and cash? Volume needs share, price, cost, working capital and capital intensity.
7. What was already expected? A good outcome can be disappointing if valuation assumed something better.

If the questions cannot be answered, the signal belongs in a watchlist, not a forecast.

Where AI helps—and where it should stop

AI can help an analyst discover relevant sources, summarize methodology, compare a new observation with a thesis and explain why a signal changed.

It should not be responsible for silently guessing a missing monthly value, choosing an entity mapping because the names look similar or converting correlation into causation.

In Altys, deterministic collectors and typed series handle numeric observations. Source links, content hashes and dates provide verification. AI operates on top of that evidence layer; it does not replace it.

The same philosophy applies across Altys research. Strategy, scorecard, research and monitoring outputs can be exported to Excel or Excel-ready formats so an investment team can challenge the calculations outside the application. Verifiability is a product requirement, not a disclaimer.

The conclusion

Alternative data can shorten the distance between a business changing and an investor noticing. It cannot eliminate the distance between activity and value.

The useful workflow is:

Observe the footprint → verify its definition and date → map it to the correct business driver → test the economics → compare with expectations → monitor what follows.

The filing may be late. The footprints do arrive every day. The advantage belongs to the investor who remembers that a footprint is evidence of movement—not proof of destination.

Explore the live public snapshots

• Alternative data for Indian stock research
• India auto sales: VAHAN and FADA
• India GST collections
• India e-way bill data
• Demat accounts and mutual-fund flows
• India insurance premium data

Data snapshot: Altys alternative-data warehouse as available on 4 September 2026. Source definitions and revision behaviour differ by series; the linked trackers provide exact publisher links and caveats. This article is educational and is not investment advice. Altys Labs is not a SEBI-registered Research Analyst or Investment Adviser.