Early access is open. Free sample datasets are live now. Download a sample

Learn

How to Choose a Market Data Provider

7 min read · Updated 19 November 2026 · By the marketdata.co data team

Every vendor's homepage says "accurate, fast, affordable". None of that tells you whether their 1-minute bars will break your backtest at the next stock split. This page is the checklist we use when we evaluate a data source, plus a short script that tests a sample file before you pay for anything.

Side by side

What actually matters when you compare vendors

"Best market data provider" depends on the job. A swing trader pulling daily bars for 500 stocks has different needs from someone replaying options quotes. Start by writing down four things: the asset classes, the smallest bar size or tick type you need, how many years of history, and whether you need it live or only historical.

Then check each vendor against these questions:

  • History start and gaps. Ask for the first date per symbol, not the marketing claim. Then look for missing days.
  • Adjustments. Are prices split adjusted, dividend adjusted, both, or neither? Can you get the raw prices too? You want unadjusted_close next to adjusted_close, not one or the other.
  • Delisted symbols. If dead companies are missing, every equity backtest you run is flattered. See our page on survivorship bias.
  • Timestamps. UTC or exchange time? Bar open or bar close? A one-bar shift is enough to leak future data into a signal.
  • Licence. Personal use and commercial use are usually priced very differently. Displaying data in an app you sell is a third category again.
  • Ownership of downloads. Some vendors make you delete data when you cancel. Others sell files you keep.

A worked example (illustrative numbers)

Say you need 10 years of 1-minute bars for 50 US stocks. Vendor A charges $99 a month with a 2-year history cap, so it fails on history before price even matters. By contrast, vendor B sells the files once at, for example, $8 per symbol, so $400 total, and you keep them. A third vendor, C, has the history but returns bars stamped at bar close while your code assumes bar open. That one is fixable, but only if you notice. The cheapest-looking option and the cheapest real option are often different vendors.

Sample data

Test a sample file before you buy

Ask every vendor for a sample in the exact format you will get. Then run checks like the ones below. This example uses our free BTC-USDT 1-minute file, which has 500 rows from the last 500 minutes of October 2026.

"""Example: sanity checks on a 1-minute bar file."""
import pandas as pd

df = pd.read_csv("btc-usdt-1m-bars.csv", parse_dates=["timestamp"])

print("1. timestamps sorted, unique, evenly spaced")
assert df["timestamp"].is_monotonic_increasing
assert df["timestamp"].is_unique
gaps = df["timestamp"].diff().value_counts()
print(gaps)  # expect a single value: 0 days 00:01:00

print("2. OHLC consistency")
bad = df[(df["high"] < df[["open", "close"]].max(axis=1)) |
         (df["low"] > df[["open", "close"]].min(axis=1))]
print(len(bad), "bars with impossible OHLC")

print("3. non-positive prices and negative volume")
print((df[["open", "high", "low", "close"]] <= 0).sum().sum(), "non-positive prices")
print((df["volume"] < 0).sum(), "negative volumes")

Read the output this way. One gap size means no missing minutes. Any bar with impossible OHLC means the vendor built bars from bad ticks or merged sessions badly. For equities, add a fourth check: pull a stock with a known split and look for the price cliff on the split date in the adjusted series. If the cliff is there, the "adjusted" data isn't.

Watch out

Common mistakes

Choosing on price per month alone. A cheap plan with a 2-year cap means you pay again somewhere else for the history. Work out the cost of the full dataset you need, not the plan.

Ignoring the licence. A personal-use licence does not cover a dashboard you charge for. Fixing that after launch is far more expensive than asking before.

Testing on a famous symbol only. Every vendor gets AAPL right. Test a small cap that was delisted, a stock that changed its ticker, and a day with a trading halt.

Mixing sources without checking time zones. Two clean files can still misalign by an hour around daylight saving changes if one is in UTC and one is in New York time.

Details

Get the data

Try our files before you trust them. We publish free samples in the same schema as our paid data: UTC timestamps, numbers as numbers, a symbol on every row.

We are in early access. Plans start at Free ($0), then Starter $29/mo, Pro $79/mo and Business $249/mo. See pricing.

FAQ

Frequently asked questions

What is the best market data provider for backtesting?

The best one is the vendor whose history, adjustments and delisted coverage match your test, at a licence you can use. For daily equity strategies that means split and dividend adjusted bars that include delisted stocks. Intraday work needs timestamps and bar alignment first. Price comes after those checks, not before.

Is free market data good enough?

Free data is fine for learning and for daily bars on large, liquid symbols. It usually falls short on delisted stocks, intraday history, and licence terms for anything you publish. Test it with the same script you would use on a paid vendor.

Should I pick one vendor for everything?

One vendor is simpler, since one schema and one set of timestamps means less glue code. In practice many traders use two, one for history and one for live data. If you do that, check that symbols and time zones line up between them.

What should I ask a vendor before paying?

Ask for the first date per symbol, whether delisted symbols are included, how prices are adjusted, what time zone timestamps use, and what the licence allows. Then ask for a sample in the exact delivery format.

Keep reading