Reference data is the descriptive and contractual information used to identify financial instruments and interpret their records: identifiers, currencies, contract sizes, expiry dates, calendars and related mappings. A price can be accurate while the resulting position or cash-flow calculation is wrong because the system attached the wrong definition to it.
Reference data tells a quant what is being measured and under which conventions. Market data supplies observations such as trades and bid–ask quotes. The distinction concerns the role of a field, not the file or feed carrying it: an exchange feed can carry instrument definitions alongside prices.
The definition beside the price
| Reference field | What a calculation needs to know |
|---|---|
| Instrument identity and identifier mappings | Which security, share class, venue or contract a vendor code denotes |
| Quote and settlement currencies | Which currency measures a price, and whether settlement uses another |
| Contract size, multiplier and units | How a count of contracts maps into underlying units or currency per price point |
| Expiry, delivery and settlement terms | Which dated obligation the position represents |
| Calendar and date conventions | Which days and rules govern trading, accrual or settlement |
| Corporate-action mappings | How old and new instrument records relate through events such as a split or merger |
A ticker symbol is useful for display but ticker reuse and venue-specific meanings make it insufficient as a universal, permanent key. A security master connects source-specific identifiers to controlled internal identities. Preserve the identifier's source, venue or scope and applicable dates; separately identify the issuer, security and traded contract where those are different objects.
Field names also need definitions. CME's futures specification distinguishes price currency from settlement currency, and specifies contract size together with a unit of measure. Its field named
ContractMultiplier has a particular meaning for variable-quantity products. A generic column called “multiplier” should therefore never be mapped to an exchange field by its name alone. CME's security-definition documentation is an example of why the source schema matters.Two contracts, but how many units?
Consider a fictional contract X whose price is quoted in USD per underlying unit. Its applicable definition says that each contract represents 100 underlying units. A historical long position contains 2 contracts:
The contract unit cancels. The result is an underlying quantity, not a USD market value, premium or margin requirement. Converting that quantity into a monetary measure requires the relevant price and the instrument's valuation or cash-flow rule. For an option, it is also not automatically the option's delta exposure.
Now suppose the fictional definition is amended to 50 units per contract from 1 June 2025. The amendment is published on 25 May and becomes usable in this research system on 26 May. The old definition was already published and usable by 1 May. We are reconstructing a position on 20 May, not valuing a position after the amendment. The contract counts used here are separate sizing lookups, not a carried-position loss calculation or a model of a real contract adjustment.
Original synthetic definition history. Dates are calendar dates in one common fictional market timezone; starts are inclusive and ends exclusive. Version A was published and usable by 1 May. USD is the quote currency, with prices per underlying unit. No post-amendment position or corporate-action cash flow is inferred.
Joining the May position to the latest record would produce 100 units instead of 200. A price series cannot repair this mistake: it originates in the units attached to the position.
Keep effective time and knowledge time
An effective date says when a definition applies. A publication time says when its source released it. A usable time records when the research or trading system could actually use it, after receipt and any required processing. An ingestion timestamp can be useful evidence, but does not by itself establish all three.
In the example, version B is unknown to this system on 20 May. On 27 May it is known, but is still not effective: a separate sizing lookup for 2 contracts on 27 May also uses 100 units per contract and returns 200 units. Both tests matter. For an interval-based record, a historical join should check that the position date falls inside the record's effective interval and that the version was available by the decision being reconstructed. A corrected record can even arrive later with a retroactive effective date; retaining the correction's availability prevents look-ahead bias.
Nasdaq's Symbol Lookup describes its information as current-trading-day data. Its directory definitions separately document file-generation timestamps and effective-date fields for particular files. Downloading a current directory does not, by itself, reconstruct the definitions available in an old backtest.
A usable reference-data process retains source records and versions, rejects unresolved or ambiguous mappings, and checks units and currencies before joining them to trades or positions. Overlapping effective intervals or missing definitions need an explicit resolution policy rather than a silent “take the latest” fallback. Reference data can change; the goal is to preserve the right definition for the question being asked.
The Instrument Data Conventions lesson connects identifiers and contract definitions to day-count, currency and pricing conventions. The public page is a preview; the full lesson requires membership.
Sources
- CME Group, MDP 3.0 — Market Data Security Definition — Futures: instrument identity, currencies, units and contract-size fields.
- Nasdaq Trader, Symbol Directory Definitions and Symbol Lookup: directory scope, field meanings and timestamp distinctions.
- Rishi K. Narang, Inside the Black Box, second edition (Wiley, 2013), Chapter 8, pp. 137–139: data-source integration, security masters and data quality.