Level 1, Level 2 and Level 3 Data
The three standard tiers of market-data detail an exchange sells — from just the best price to the full order book to every individual order — and why more detail costs more and demands more infrastructure to consume.
Prerequisites: Order Book Mechanics
Ask "what's the price of this stock right now" and there are several genuinely different answers depending on how much detail you're willing to pay for and process. Exchanges package market data into standard tiers, conventionally called Level 1, Level 2, and Level 3, each layering on more of the order book's internal structure — and each demanding proportionally more bandwidth, storage, and processing power from anyone consuming it.
Level 1 is the shallowest and most widely available: just the best bid, the best ask, and the last trade price and size — the National Best Bid and Offer, in US equities terms. It answers "what would I pay or receive right now" and nothing more; you can't see how much size sits behind that price, or what's queued up behind it. Level 2 exposes the full depth of book: every price level currently resting in the order book, with the aggregate size at each level, letting you see not just the best price but how much liquidity would need to be consumed to move the price further, and how it's distributed across levels above and below the touch. Level 3 goes one step further, exposing individual, order-level detail — every distinct resting order, often with its own ID, letting you reconstruct precisely which order is where in the queue, when it arrived, and when it's modified or cancelled, rather than just an aggregated total per price level.
Worked example: what each tier can and can't tell you
The order book shows $100.00 bid for 5,000 shares in aggregate, $100.05 offered for 3,000 shares in aggregate. A Level 1 feed shows exactly two numbers: best bid $100.00, best ask $100.05 (plus the last trade) — it cannot tell you the 5,000 and 3,000 figures at all, so you have no way to gauge whether the market can absorb a 2,000-share order without moving. A Level 2 feed shows those aggregate sizes at every price level, so you can see the $100.00 level has 5,000 shares and, two ticks down, $99.90 has another 8,000 — enough to estimate how far a large order would push the price. A Level 3 feed additionally shows that the 5,000 shares at $100.00 is actually five separate resting orders of 1,000 each, placed at different times — letting you compute your exact queue position if you placed an order behind them, something Level 2's aggregate number alone cannot reveal.
What this means in practice
Level 1 is sufficient for most retail and even many systematic strategies that just need a fair current price. Level 2 is the standard requirement for anything that needs to reason about liquidity and market impact — most execution algorithms and market-making strategies. Level 3 is mainly needed by strategies that model queue position directly (deciding whether an order is likely to fill soon based on what's ahead of it) and comes with a much larger data-processing burden: reconstructing and maintaining an accurate, order-by-order book from a Level 3 feed in real time is a nontrivial systems problem in its own right.
Level 1 gives only the best bid/ask and last trade; Level 2 adds aggregate size at every price level (full depth of book); Level 3 breaks each level down into individual, identifiable orders. Each tier costs more bandwidth and processing than the last, and the right choice depends on what the strategy actually needs to see.
If a strategy's logic never references anything besides "the best price" and "did a trade just happen," Level 1 is enough — don't build infrastructure to consume and reconstruct a Level 2 or Level 3 book you never actually query.
Related concepts
Practice in interviews
Further reading
- Harris, Trading and Exchanges, ch. 3