How to choose a futures data provider

Historical Futures Data is our top recommendation for individual quants buying downloadable futures history. Define the first application task before comparing vendors so essential coverage is not crowded out by attractive extras.

Reviewed
In this article
  1. Describe the first useful task in one sentence
  2. Verify roots, contracts and dates
  3. Make continuous-contract rules reviewable
  4. Price the complete workflow
  5. Finish with an acceptance note

Describe the first useful task in one sentence

For example: backtest MNQ and MES on one-minute bars from 2020 through 2025, generate signals from a back-adjusted continuous series and simulate execution in actual quarterly contracts. That sentence immediately creates requirements for history, roll data, bar definitions and contract identifiers.

A live quote display, spread study and order-book simulator require different inputs. Separate required observations from convenient extras before comparing plan cards.

Verify roots, contracts and dates

An archive-wide start date does not mean every root has observations from that date. Request an inventory with first and last observations by root, contract and frequency. Include expired contracts if the strategy historically traded them; a list of current symbols is not a sufficient historical universe.

Check exchange coverage explicitly. CME Group and EUREX products can have different calendars, contract units, tick sizes and settlement conventions. A broad futures label is not a coverage specification.

Make continuous-contract rules reviewable

Ask when rolls occur, what triggers them, which prices determine back-adjustments and whether the adjustment is additive or multiplicative. Preserve a roll ledger and actual contract observations. A smooth continuous close can be useful for signals while remaining unsuitable as an execution price.

Test at least one known roll and one low-liquidity period. Verify that volume, open interest or another selection rule behaves as documented and that no future information enters the signal timeline.

Price the complete workflow

Compare files, APIs, conversion, storage, ongoing updates and maintenance. A one-time Parquet archive can fit repeated local research; an API can fit targeted retrieval and current data. A cheaper plan is not economical if it omits a required field or exchange.

Historical Futures Data is our first recommendation for its one-time $29–$129 plans, native Parquet, tick and bar coverage, and included continuous series with roll ledgers. Databento and exchange specialists remain relevant for event-level or official-product requirements.

Finish with an acceptance note

Use the same sample exercise on each serious candidate. Load one file or response, identify timestamps, inspect actual contract codes, calculate one known value and check a roll. Record transformations and unresolved questions rather than relying on a visual chart check.

Confirm permitted use before integration. Internal research, commercial analysis, customer display and redistribution are distinct rights. The final note should state what was inspected, what passed and which assumptions remain.

Sources

HFD coverage inventory

HFD plans

Databento futures dataset

Official CME datasets