MarketParquet vs Databento
> exchange-direct real-time + historical data, usage-priced
Databento is a premium, exchange-direct data platform: full order-book and tick fidelity, real-time streaming, and usage-based pricing, aimed at HFT and microstructure research. MarketParquet is the opposite end -- flat-fee, end-of-day-to-1-minute bars as Parquet files for backtesting. They solve different problems; here's how to tell which one you actually need.
Side-by-side
| Databento | MarketParquet | |
|---|---|---|
| Asset classes | US equities, futures, options | US stocks, ETFs, futures |
| Data granularity | full order book (MBO/L3), tick, 1s-1d bars | 1-minute to daily bars |
| Equity history depth | tick from 2018; daily backfill toward ~2010 | stocks/ETFs from 2000, futures from Dec 2007 |
| Format | DBN binary; CSV/JSON/Parquet export via client | Apache Parquet files, by-date partitioned |
| Real-time data | yes (~590us latency) | no -- end-of-day batch only |
| Price adjustments | raw/unadjusted + separate corporate-actions API | split-adjusted OHLCV out of the box |
| Pricing model | usage-based per GB, or $199+/mo subscription | one-time $79, no metering |
| Free tier | $125 signup credit (expires) | recent daily sample, unlimited, no card |
| Includes delisted tickers | yes | yes |
Where Databento wins
- Exchange-faithful, event-level fidelity: full order book (MBO/L3), tick-by-tick, nanosecond timestamps.
- Real-time streaming with HFT-grade (~590us) latency -- MarketParquet is end-of-day only.
- Options and deep futures microstructure across many venues.
- Genuine pay-as-you-go: buy exactly the symbols, dates, and schema you need, plus a $125 signup credit.
Where MarketParquet wins
- Flat $79 one-time for the full archive -- no per-GB metering, no four-figure monthly license.
- Split- and dividend-adjusted OHLCV out of the box; no wiring up a corporate-actions API to back-adjust yourself.
- Deeper equity history for backtesting: 1-minute-to-daily bars from 2000, vs Databento's US-equity tick starting 2018.
- Plain Parquet for pandas/polars/duckdb -- no DBN tooling or streaming client to learn.
Which to pick
Pick Databento if you need real-time feeds, full order-book (MBO/L3) or tick data, options, or exchange-faithful microstructure for HFT research.
Pick MarketParquet if you're backtesting US equities, ETFs, or futures on 1-minute-to-daily bars and want a flat one-time price and ready-adjusted Parquet instead of metered, raw event data.
FAQ
Is Databento cheaper than MarketParquet?
It depends how much data you pull. Databento is usage-based (priced per GB) with a $125 signup credit, or subscriptions from $199/month; MarketParquet is a flat $79 one-time for the complete archive. For deep, wide historical backtests a flat fee avoids per-GB surprises; for a small, targeted slice of tick data, Databento's pay-as-you-go can be cheaper.
Does Databento provide Parquet?
Yes -- Databento's native format is DBN, and its Python client can transcode to Parquet, CSV, or JSON. MarketParquet ships Parquet directly, partitioned by trading day, with no client library or transcoding step required.
Is Databento's data split-adjusted?
No -- Databento delivers raw, exchange-faithful prices plus a separate corporate-actions/adjustment-factor API you use to back-adjust yourself. MarketParquet delivers split-adjusted OHLCV out of the box, which is simpler if you just want ready-to-backtest bars.
How far back does Databento's history go?
Databento's US-equity tick and order-book history starts in May 2018 (CME futures go back to 2010, with daily backfill in progress). MarketParquet provides 1-minute-to-daily US equity/ETF history from 2000 and futures from December 2007.
> Pricing and feature snapshots reflect public information at time of writing and may shift. Always confirm against Databento's own pricing page before committing.