Knowing how to build a correct pipeline is only half the job; knowing where the raw data actually comes from is the other half, and it is a genuinely different skill, evaluating a vendor's coverage, cost, license terms, and adjustment/PIT/survivorship guarantees before you have written a single line of pipeline code against it. What follows is a deliberately concrete, named map of both markets, split first by whether the source is paid (institutional-grade, typically with contractual PIT and survivorship guarantees built in) or free (usable immediately, typically with more manual reconciliation work left to you), and separately called out where a source's specific selling point is exactly this module's topic, being survivorship-bias-free or point-in-time-correct out of the box.
Which US financial data providers are paid?
United States: paid, institutional. Bloomberg (Terminal, and the programmatic Bloomberg Data License / B-PIPE feed) and LSEG Workspace (the London Stock Exchange Group's current branding for the platform long known as Refinitiv Eikon, itself the successor to Thomson Reuters' data business) are the two dominant full-service institutional terminals, covering real-time and historical prices, fundamentals, news, and analytics globally, at a cost that puts them squarely in "your employer pays for this" territory rather than "an individual quant buys this." FactSet and S&P Capital IQ compete in the same institutional tier, particularly strong on fundamentals and estimates. ICE Data Services and Morningstar Direct round out the large full-service names. For a working quant developer without an institutional budget, the more accessible paid tier includes Polygon.io and Databento (both API-first, real-time and historical, priced for individual developers and small funds), Tiingo (a genuinely affordable freemium/subscription API with clean adjusted EOD and fundamentals data), and Nasdaq Data Link (the platform formerly branded Quandl, a marketplace of both free and paid datasets from many providers). For the specific concern this module drills hardest, survivorship. Norgate Data is the name to know: a specialist vendor whose core selling point is explicitly survivorship-bias-free historical equities data (delisted names included, by design), popular precisely among quant developers who have learned this module's lesson the hard way. In academic and institutional research specifically, CRSP (typically accessed via WRDS, Wharton Research Data Services) paired with Compustat for fundamentals is the long-standing gold-standard combination, prized for exactly the same survivorship-bias-free, point-in-time-correct guarantees this module teaches you to build by hand.
Advertisement
Which US financial data sources are free?
United States: free / official. SEC EDGAR (data.sec.gov, including the companyfacts XBRL API this program's own tools/fetch_edgar.py already talks to) is the single best free source for US company fundamentals, it is the primary filing repository itself, public domain, and as authoritative a source as exists; note that EDGAR gives you the filed documents with their real filing dates, which makes it a genuinely excellent free source for point-in-time-correct fundamentals if you build the PIT logic yourself (exactly what point-in-time discipline does). FRED (Federal Reserve Economic Data, from the St. Louis Fed) is the standard free source for US macroeconomic series. Stooq publishes free, simply-formatted historical EOD price CSVs. Yahoo Finance remains a widely-used free historical-price source (commonly accessed programmatically via the unofficial yfinance library) but carries a real, worth-knowing caveat: it is not an official data product, its Terms of Service around programmatic/bulk use have shifted before and can shift again, and its adjustment methodology, while generally solid, is not contractually guaranteed the way a paid vendor's is, treat it as a fine source for learning and prototyping, and verify independently before anything that matters rides on it. Alpha Vantage offers a free API tier (rate-limited) alongside its paid one.
Which Indian financial data providers are paid?
India: paid, institutional. Bloomberg and LSEG Workspace both operate full India coverage as part of their global platforms. India-specific institutional vendors include CMIE (Centre for Monitoring Indian Economy), whose Prowess database of corporate financials and Economic Outlook macro database are the closest Indian equivalent to a Compustat/FRED combination, long-standing and widely used in Indian academic and institutional research; Capitaline (Capital Market Publishers), a long-established Indian financial database covering listed-company fundamentals and prices; and ACE Equity (from Accord Fintech), a similarly well-established Indian corporate and market database. For real-time and historical India market data at a more developer-accessible price point, TrueData and Global Data Feed (GDFL) are established Indian vendors serving the algorithmic-trading community, and Zerodha's Kite Connect is a broker API (not a pure data vendor in the institutional sense, but worth naming because it is the most common way an individual Indian quant developer actually gets live and historical NSE/BSE data programmatically today). NSE and BSE themselves also sell bulk historical and real-time data products directly to institutional subscribers.
Which Indian financial data sources are free?
India: free / official. NSE and BSE both publish a daily bhavcopy, the official end-of-day price-and-volume dump for every listed security, freely on their own websites, and it is as authoritative a free EOD source as exists for Indian equities (this is the real-world analogue this program's bundled, synthetic nse-eod dataset stands in for). AMFI (the Association of Mutual Funds in India) publishes daily mutual-fund NAV data freely. The RBI's own DBIE (Database on Indian Economy) is the free, official source for Indian macroeconomic and monetary data, the direct analogue of FRED for India. data.gov.in is India's open-government-data portal, useful for a wider range of public datasets beyond markets specifically. For fundamentals and screening, Screener.in is an extremely popular, genuinely useful free (with a paid tier for advanced features) source for Indian company financials, widely used by the retail and prosumer investing community; Trendlyne offers a comparable freemium product. Company filings themselves are also available directly and freely from the MCA (Ministry of Corporate Affairs) and from each exchange's own filings section.
How do you match a need to a source?
Matching a need to a source. As a working rule of thumb: if your employer or institution already pays for Bloomberg, LSEG Workspace, FactSet, or a CMIE/Capitaline/ACE Equity subscription, use it, the PIT and survivorship guarantees are typically contractual, and reinventing this module's logic on top of a feed that already handles it correctly is wasted effort (though understanding the logic, which is the entire point of this module, is what lets you check the vendor's claim rather than take it on faith). If you are building on a budget, Polygon.io/Databento/Tiingo (US) and TrueData/GDFL/Kite Connect (India) give you clean, developer-friendly APIs, but you should independently verify their adjustment methodology and whether delisted names are actually retained (many are not, by default), this is precisely where this module's own flag_unadjusted_jumps and the discipline of building your own historical-membership record earn their keep. If you are learning, prototyping, or building this program's own exercises, SEC EDGAR + FRED + Stooq (US) and NSE/BSE bhavcopy + RBI + Screener.in (India) cost nothing and, combined with this module's adjustment and PIT logic built by hand, get you further than most beginners expect. And if survivorship-bias-free history specifically is the one thing you cannot compromise on for a serious backtest, Norgate Data or a CRSP/WRDS academic subscription are the names built to guarantee exactly that.
{
"skill": "Researching and comparing data-vendor documentation quickly, without treating a vendor's marketing claims as verified fact",
"use": "Use an AI assistant to summarize a vendor's stated data coverage, adjustment methodology, and pricing tier from their public documentation — a genuinely useful first-pass research accelerant — while treating every specific claim (does this feed retain delisted names? is 'adjusted close' split-only or split-and-dividend?) as unverified until you find it stated on the vendor's own current page.",
"tools": ["Chat assistants — Claude, ChatGPT, Gemini", "A research/browsing-enabled assistant that can actually open the vendor's current documentation page"],
"prompt": "I'm evaluating <VENDOR> for <MY USE CASE, e.g. 'a survivorship-bias-free US equities backtest'>. Summarize, from their public documentation: (1) whether their historical data retains delisted/acquired companies, (2) exactly what 'adjusted close' means in their feed (splits only, or splits and dividends), (3) their fundamentals filing-lag/point-in-time convention if any, (4) their current pricing tier for an individual developer. Flag anything you could not find a direct documentation source for.",
"verify": "Vendor names, pricing, and even product names are perishable — a platform can be rebranded (LSEG Workspace was, not long ago, Refinitiv Eikon), a free tier can shrink, a company can be acquired. Open the vendor's OWN current page for every claim before acting on it, especially pricing and the survivorship/PIT guarantees this module cares about most; a model's training-data snapshot of a vendor's offering can be stale in exactly the way AI0.01 warns about. Never select a data vendor for a real strategy based on an AI summary alone.",
"diy": "You must already know, from this module, exactly WHICH properties to check (adjustment convention, delisted-name retention, filing-lag/PIT convention) — without that checklist from Sections 2-8, a vendor-comparison prompt just produces a plausible-sounding summary of features that don't actually address the risks this module teaches you to worry about.",
"market": "both"
}