Skip to content

ISIN CUSIP SEDOL Cross Reference Service

isin-cusip-sedol-cross-reference-servicesource

Use when building a security master or validating symbols at order entry, checking ISIN, CUSIP, SEDOL and FIGI check digits and resolving any one of them to the others. A valid check digit is not proof the security exists.

Domain
Data Management
Version
2.0.0
Reading
6 min
Hands off to
5
Handed off from
8
License
Apache-2.0
CoversOpenFIGI APICUSIP Global ServicesLSE SEDOL MasterfileANSI X9.145 FIGIPython Dataclasses

When to Use

Use this skill when building a Security Master, ingesting multi-vendor reference data (Bloomberg, Refinitiv, FactSet), or validating symbols at order entry. Global securities carry disparate identifiers — ISIN (12 characters, global, ISO 6166), CUSIP (9 characters, North America, ANSI X9.6), SEDOL (7 characters, LSE), and FIGI (12 characters, ANSI X9.145) — and a typo in any of them silently corrupts every downstream join keyed on it.

The four are all "Modulo 10" schemes, and that description is the trap: the algorithms are different and are not interchangeable.

Identifier Algorithm Detail
ISIN Luhn over the expanded digit string Each character becomes its value first (A=10 … Z=35, so a letter yields two digits), then Luhn right-to-left; total $\bmod 10 = 0$
CUSIP Double-add-double over characters 1–8 1-indexed even positions doubled; check digit $= (10 - \text{sum} \bmod 10) \bmod 10$
SEDOL Weighted sum, weights $(1, 3, 1, 7, 3, 9)$ Nothing is doubled. Not Luhn
FIGI Double-add-double over characters 1–11 1-indexed even positions doubled — deliberately offset from ISIN by the standard

When NOT to Use

  • As proof a security exists. A check digit catches single-character substitutions and most transpositions. It says nothing about whether the identifier was ever issued, is still active, or points at the instrument you think it does. Only the issuing agency (CUSIP Global Services, the LSE SEDOL Masterfile, the local National Numbering Agency) or the OpenFIGI mapping API can answer that. This module makes no network calls.
  • As a multi-listing security master. An ISIN identifies an issue; a SEDOL identifies one security on one market. A cross-listed issue therefore has one ISIN and several SEDOLs. The flat one-record-per-security schema here models a single listing — a real multi-venue master needs a one-to-many SEDOL (and FIGI) table. See reference-data-symbol-mapping-across-vendors.
  • To infer an identifier's type when you already know it. Length-and-shape inference cannot separate a 7-character ticker from a SEDOL, or a Cayman ISIN from a FIGI. A vendor feed column is labelled; pass identifier_type= and skip the guessing entirely.
  • As a substitute for corporate-action handling. A ticker rename or share-class reorganisation changes the mapping, not the check digit. See corporate-action-event-calendar-integration and instrument-universe-change-detection-and-alerting.
  • As a CUSIP redistribution vehicle. CUSIP data is licensed from CUSIP Global Services. Validating a check digit is arithmetic on a string you already hold; publishing a CUSIP database is a licensing question, not a technical one.

Prerequisites

  • An identifier string (ISIN, CUSIP, SEDOL, FIGI or ticker), and — strongly preferred — the type the source feed says it is.
  • Security master rows as SecurityMasterRecord (isin, cusip, sedol, figi, ticker_symbol, asset_name, country_code). Any of the four structured identifiers may be an empty string where the issue has none.
  • A decision on strict_validation: raise on bad master data (default), or load it and audit via validate_master_data().

Workflow

  1. Validate the master data before it can corrupt a join:
    • On construction the engine re-validates every identifier on every row against its own algorithm and indexes each type separately. A row whose SEDOL fails its own check digit is a data defect, not a lookup problem — it is reported by validate_master_data() and raises under strict_validation=True.
    • Offending identifiers are still indexed under strict_validation=False. Refusing to index them would make the row unreachable by a key live systems may already be joining on, which trades a visible defect for an invisible one.
    • Duplicate identifiers across rows are reported, and the first row wins. Two rows claiming one ISIN is an upstream merge error; resolving it silently hides it.
  2. Normalise, then classify by syntax — never by length alone:
    • Uppercase, strip surrounding whitespace, and strip embedded spaces and hyphens for structured matching only (vendors ship CUSIPs as 037833-10-0). Ticker matching uses the string as typed, because - and . are meaningful in ticker symbols (BRK.B).
    • Test the normalised string against each identifier's full syntax rule, then its check digit. A string can satisfy more than one: KYG875721634 (Tencent Holdings) is a valid ISIN and satisfies the FIGI syntax rules and passes the FIGI check digit. Report every type that validated in candidate_types rather than silently choosing.
    • When more than one type validates, prefer whichever one the master data actually resolves — that is evidence. Fall back to the documented CLASSIFICATION_PRIORITY only when the data cannot break the tie.
  3. Separate "corrupted" from "unknown":
    • A string shaped like a structured identifier whose check digit fails is INVALID_CHECKSUM — reject it at the boundary, do not attempt a lookup.
    • A string matching no structured syntax falls through to TICKER and is looked up without a check digit. The report sets checksum_applied=False so is_checksum_valid=True cannot be misread as "verified".
  4. Resolve against a type-scoped index:
    • Look the query up in the index for its resolved type only. Searching the union of all five fields lets a query validated as a CUSIP resolve on some unrelated row's ticker.
  5. Emit IdentifierCrossReferenceReport carrying query_type, candidate_types, is_checksum_valid, checksum_applied, status and audit_notes — enough for a reviewer to see not just the answer but how ambiguous the question was.

Full procedure: see references/workflows.md. Standards reference: see references/standards.md. Printable pre-flight checklist: see assets/checklist.md.

Common Pitfalls

  • Running one algorithm against another identifier. SEDOL is a plain weighted sum; applying Luhn to it accepts corrupted SEDOLs. FIGI's doubling is offset from ISIN's by design — ANSI X9.145 says so explicitly, so that the same string generally gets a different check digit under the two schemes. Reusing the ISIN routine for FIGI is a silent correctness failure.
  • Admitting vowels into a SEDOL. Vowels are never used, so the character set is 0-9 plus the consonants BCDFGHJKLMNPQRSTVWXYZ. A permissive [B-Z0-9] class admits E, I, O and U — and I-for-1 and O-for-0 are the exact typos the excluded-vowel rule exists to catch.
  • Assuming 12 characters means ISIN. A FIGI is also 12 characters, also starts with two letters, and also ends in a digit. Classifying on length alone routes every FIGI into the ISIN validator, where it fails the check digit and gets rejected as corrupt.
  • Trusting a check digit as an existence proof. BBG000MM82B1 passes the FIGI check digit perfectly and is not Meta's FIGI. Arithmetic validates typing; only the issuing agency validates existence.
  • Assuming a US ISIN equals its CUSIP. A US ISIN is US + the 9-character CUSIP + a newly computed check digit (037833100 → US0378331005). The CUSIP's own check digit is not reusable. Note also that a Private Placement Number CUSIP containing *, @ or # cannot be embedded in an ISIN at all.
  • Calling int() on a check digit position the regex never constrained. A CUSIP pattern of [A-Z0-9]{9} admits a trailing letter, and the subsequent int() raises instead of returning False — one malformed vendor row kills the whole ingest loop. Constrain the check-digit position in the pattern.
  • Keying the master on ticker rather than FIGI. Tickers are reassigned and renamed; FB became META while the FIGI BBG000MM2P62 did not move. Join on the FIGI and carry the ticker as an attribute.
  • Treating one SEDOL as the security. A cross-listed issue has one ISIN and a SEDOL per market. A UNIQUE constraint on a single SEDOL column silently drops listings.

Verification

  • Instantiate IsinCusipSedolCrossReferenceEngine(). Confirm validate_master_data() returns () — every shipped row's ISIN, CUSIP, SEDOL and FIGI must survive its own validator, and each must resolve back to its own row.
  • Look Apple Inc up by all five keys — ISIN US0378331005, CUSIP 037833100, SEDOL 2046251, FIGI BBG000B9XRY4, ticker AAPL $\implies$ each returns MATCH_FOUND with the matching query_type. Look up 037833-10-0 $\implies$ the hyphenated CUSIP normalises and still resolves.
  • Audit a corrupted ISIN (US0378331009) $\implies$ status == "INVALID_CHECKSUM", query_type == "ISIN", is_checksum_valid false, no record returned.
  • Verify the ambiguity path: look up KYG875721634 $\implies$ candidate_types == ("ISIN", "FIGI") and audit_notes flags AMBIGUOUS.
  • Verify the FIGI implementation against the worked example carried in ANSI X9.145-2021 itself: NRG92C84SB39 $\implies$ valid, and every other check digit for that 11-character stem $\implies$ invalid.
  • Verify master-data enforcement: construct a record carrying SEDOL BNPYS71 $\implies$ ValueError under strict_validation=True, and a single reported problem naming the SEDOL under strict_validation=False.
  • Run python -m unittest discover -s skills/isin-cusip-sedol-cross-reference-service/scripts.

Verify it, from the repository root

python -m unittest discover -s skills/isin-cusip-sedol-cross-reference-service/scripts

Hands off to 5

Skills this document names, usually in When NOT to Use, as the owner of a case it excludes.

Calendars, timezones, symbology

Handed off from 8

Skills that name this one as the place a case belongs. The reverse edges of the graph.

Calendars, timezones, symbology

Leakage-free features and models

SEC, FINRA, MiFID II, FCA, SEBI