In this Article
Market data looks like the most available data in the world: prices are on every finance site. It is also one of the few categories where the legal and commercial structure, not the technical access, decides what you can build.
This guide explains how exchange licensing works, what delayed data means, the four provider tiers and what each is for, and why the usual scrape-it-yourself instinct produces both legal and quality problems in this specific domain.
Key Facts
- Exchanges license market data, they do not simply sell it. Redistribution, display and derived use are separate permissions with separate fees.
- Delayed data is a licensing category, not a technical delay. The common fifteen-minute figure exists because that is what exchanges permit without a real-time agreement.
- Scraping quotes is the wrong shortcut here. Unlike most public web data, market data carries explicit contractual terms, and a broker page is a licensee, not a source.
- Corporate actions are where free data breaks. Splits, dividends and symbol changes silently corrupt any history that does not adjust for them.
- Survivorship bias ruins backtests. A dataset containing only companies that still exist will make almost any strategy look profitable.
Why is market data licensed rather than sold?
Because exchanges treat the prices generated on their venues as a product, and they charge separately for each kind of use.
The permissions are typically split between internal use, external display, redistribution and derived works, and the fees differ by whether users are professional or non-professional. A provider selling you an API has its own agreement with the exchanges, and your access is a sublicence under it. That is why the terms of a market data API constrain what your application may do with the numbers, not just how many calls you may make.
The practical consequence is that the question to answer before choosing a provider is not what it costs, but what your product does with the data: showing a price to logged-in users, storing history, computing a derived indicator and publishing it are four different permissions.
What do the provider tiers actually give you?
Every stock market data api sits in one of four tiers, and picking the wrong one is the most common and most expensive mistake. We call it the 4-layer market data stack.
| Tier | What you get | Fits |
|---|---|---|
| 1. Free and delayed | End-of-day or delayed quotes, limited history, tight rate limits | Prototypes, personal tools, research where latency is irrelevant |
| 2. Retail API | Delayed or limited real-time, clean symbology, documented corporate actions | Most applications, dashboards, alerting |
| 3. Professional feed | Real-time consolidated data, full depth, licensed redistribution | Trading products and anything customer-facing at scale |
| 4. Direct exchange | Raw venue feeds, lowest latency, heaviest obligations | Latency-sensitive trading only |
Teams routinely prototype on tier one and discover at launch that their use requires tier three, with a price and a compliance process to match. Deciding the eventual tier at design time is cheaper than migrating a product’s data layer later.
Why is scraping the wrong approach here?
Three reasons, and they are stronger than the usual arguments about public web data.
The terms are explicit. Finance sites and broker platforms display data under their own exchange agreements, which almost universally prohibit redistribution and systematic extraction. Taking numbers from a licensee does not give you the licence.
The quality is not there. Displayed prices are frequently delayed, sometimes from a different venue, rounded, and rarely adjusted for corporate actions. A price history stitched from a web page will be wrong at exactly the points that matter: splits, dividends and symbol changes.
The failure mode is invisible. A scraper that misses a split does not error, it produces a chart with a cliff in it, and any model trained on that history learns the artefact.
Where ordinary collection infrastructure legitimately applies in finance is the non-price data around the market: filings from official regulators, company announcements, public procurement, news and job postings. Those are public documents and are what DataImpulse residential is for, at $1 per GB across 195 countries. Prices are not that category.
What breaks a historical dataset?
Five problems, and each has a clear test. Use the check when the condition in the middle column applies, and avoid assuming a provider handles it silently.
| Problem | Use this check when | Symptom if ignored |
|---|---|---|
| Corporate actions | Any history longer than a few months | Impossible price gaps on split dates |
| Survivorship bias | Any backtest | Almost every strategy appears profitable |
| Symbol reuse | Long histories across delistings | Two unrelated companies merged into one series |
| Venue differences | Comparing sources | Small persistent discrepancies read as data errors |
| Timezone and session handling | Intraday work | Bars misaligned by hours around daylight saving changes |
Ask any provider directly how they handle each row. Adjusted history and a delisted-securities archive are the two things that separate a research-grade dataset from a display feed, and they are rarely visible on a pricing page.
Frequently Asked Questions
Why is real-time market data expensive?
Because exchanges license it rather than sell it, and charge separately for internal use, display, redistribution and derived works, with different rates for professional and non-professional users. A provider’s API price includes its own sublicence obligations to those exchanges.
What does delayed market data mean?
It is a licensing category rather than a technical limitation. Exchanges permit distribution of quotes after a defined interval, commonly fifteen minutes, without a real-time agreement, which is why free tiers converge on the same delay.
Can I scrape stock prices instead of paying for an API?
It is the wrong shortcut in this specific domain. Finance sites display prices under their own exchange agreements that prohibit redistribution and systematic extraction, and the displayed numbers are delayed, rounded and unadjusted for corporate actions, so the data is both contractually and technically unsuitable.
What ruins a backtest most often?
Survivorship bias and unadjusted corporate actions. A dataset containing only companies that still exist makes nearly any strategy look profitable, and unadjusted prices create artificial cliffs on split dates that models learn as signals.
Where do proxies legitimately fit in financial data?
Around the market rather than in it: regulatory filings, company announcements, public procurement, news and hiring data are public documents that render differently by region. Price data is a licensed product and belongs with a licensed provider.
The public documents around the market
Filings, announcements, procurement notices and news are public and region-dependent. DataImpulse residential proxies give country-pinned exits at $1 per GB across 195 countries. Create an account and collect the document side reliably.
Related: government contract data · is web scraping legal · proxies for web scraping.
Last updated: September 17, 2026.
State/City/Zip/ASN Targeting 
