Skip to content
FXOpen Ledger
Directory

FX Global Code Last Look Execution

FIX Tag 39 & Tag 103 Reject Code Forensics in OTC FX Aggregators

Lumping every rejected FIX ExecutionReport into a single fill-ratio KPI blinds OTC foreign exchange desks; parsing Tag 103 OrdRejReason and normalized Tag 58 reason codes separates Prime Broker NOP breaches from LP last-look tolerance rejects.

By Alistair Kerr, CFAReview not recorded
FIX Tag 39 & Tag 103 Reject Code Forensics in OTC FX Aggregators — FXOpen Ledger 2026 quantitative & supervisory data card
FXOpen Ledger (FX Global Code Principle 17, Last-Look Hold Times & Execution TCA): Separating taker limit collars, PB NOP credit halts, and true LP last-look rejects in FIX 35=8 logs.Source: FXOpen Ledger Institutional Research DeskLicense: CC-BY-4.0

When an institutional FX desk sees its weekly fill ratio on Liquidity Provider Delta drop from 93% to 74%, penalizing LP Delta on the quarterly scorecard without inspecting underlying FIX 4.4 / FIX 5.0 ExecutionReport (35=8) tags frequently misdiagnoses an internal operational failure as LP misconduct! In electronic foreign exchange aggregators, an outbound NewOrderSingle (35=D) returning Tag 39 (OrdStatus) = 8 (Rejected) or Tag 39 = 4 (Canceled - IOC Expired) can stem from six completely unrelated layers: client-side limit-price collar protection, ECN message-rate throttling, FX Prime Brokerage Designation Notice Net Open Position exhaustion, stale Tag 117 QuoteID collision, or true LP last-look price-tolerance rejection.

Embedded within our FX Global Code Principle 17 & Execution TCA cornerstone and the Principle 17 & Execution TCA hub, Tag-level FIX log normalization is the prerequisite for accurate GFXC LP cover sheet reconciliation.

Taxonomic Mapping of FIX Tag 39, Tag 103, and Tag 58 Strings#

Standard FIX protocol defines Tag 103 (OrdRejReason) integers (0 = Broker/Exchange option, 1 = Unknown symbol, 3 = Order exceeds limit, 6 = Duplicate order, 11 = Unsupported order characteristic, 13 = Incorrect quantity, 15 = Unknown account, 99 = Other), yet historically many bank gateways lazily stamped 39=8 | 103=0 or 103=99 on everything and buried the real cause inside free-text Tag 58 (Text)! Under GFXC transparency guidelines, major multi-dealer ECNs and FIX API liquidity bridges (OneZero, PrimeXM, Integral, Lucera) now normalize reject telemetry into five canonical buckets:

  1. Taker Limit-Collar Expiration (39=4 IOC Unfilled / 58=LimitPriceMissed): Client sent a Marketable Limit IOC with 0.0-pip slippage tolerance (Price=1.08502), and top-of-book ticked to 1.08503 while the packet was in flight on fiber. Zero LP fault—pure tight client slippage cap!
  2. QuoteID / MQL Race Collision (103=99 | 58=QuoteExpired_Tag117): Two takers hit the exact same firm or ZAHT QuoteID within 50 microseconds; the first taker swept the full $5M displayed size, leaving zero liquidity for the second packet.
  3. PB / PoP Credit & NOP Gate Halt (103=3 | 58=CreditLimitBreach_NOP): Prime Broker or Prime-of-Prime pre-trade risk filter blocked the ticket before the LP pricing engine ever evaluated price.
  4. Throttle / Quote-to-Trade Governor (103=99 | 58=MaxOrderRateExceeded): Client algo burst exceeded max messages per second on that session.
  5. True LP Last-Look Price-Tolerance Reject (103=0 | 58=LastLook_PriceTolerance): Conditional LP invoked Principle 17 tolerance check and declined the trade.
FIX ExecutionReport 35=8 reject classification pipeline separating credit, QuoteID, and last-look codes
FIX ExecutionReport 35=8 reject classification pipeline separating credit, QuoteID, and last-look codesVerified Exhibit

Clean LP Fill Ratio Formula vs Raw Unfiltered Fill Rate#

Evaluating market makers using unfiltered fill ratios distorts smart-order-router shadow-spread weights (derived in firm vs last-look reject economics). By excluding client-induced limit collar misses ($R{collar}$), credit breaches ($R{credit}$), and session throttles ($R{throttle}$), execution engineers compute the unbiased **Clean LP Acceptance Ratio ($p{clean}$)**:

Deriving Clean LP Acceptance Ratios From Raw Drop-Copies#

Formula or model from the articleNot independently verified
p_raw = N_filled / (N_filled + R_collar + R_credit + R_throttle + R_quote_race + R_last_look)
  • pclean = Nfilled / (Nfilled + Rlastlook + wrace * Rquoterace)
  • Where:
  • N_filled = Count of fully or partially filled tickets (OrdStatus 39=2 or 39=1)
  • R_collar = Client limit-price IOC misses where LP quote already updated prior to ingress
  • R_credit = Prime Broker / PoP NOP credit blocks (Tag 103=3)
  • R_throttle = Message-rate governor blocks
  • Rquoterace = Valid simultaneous depletion of displayed QuoteID size (weighted w_race = 0.25 to 0.50 if phantom depth suspected)
  • Rlastlook = True LP Principle 17 conditional price-tolerance rejections
  • Worked Drop-Copy Audit (10,000 Outbound USD/CAD 2M IOC Tickets on Stream Delta):
  • Nfilled = 8,160; Rcollar = 790 (client set 0.0p max slippage!); R_credit = 210 (Friday afternoon NOP cap);
  • Rthrottle = 40; Rquoterace = 130; Rlast_look = 670.
  • p_raw = 8,160 / 10,000 = 81.60% (Falsely breaches 85% router demotion threshold!)
  • p_clean = 8,160 / (8,160 + 670 + 0.5 * 130) = 8,160 / 8,895 = 91.74% (Healthy ZAHT LP; fix client 0.0p limit collar & NOP line!)

FIX 35=8 Tag Attribution & Desk Remediation Matrix#

Quantitative Distribution · Normalized Failure Root Cause
Normalized Failure Root CauseTypical RTT Latency
01`39=459=302`39=8103=303`39=8103=9904`39=8103=0
FIX Tag Signature (35=8 / 35=9)Normalized Failure Root CauseTypical RTT LatencyResponsible Desk OwnerImmediate Remediation Action
`39=459=358=IOC_NoMatch`Tight Taker Limit Collar Miss0.8 - 2.2 msBuy-Side Algo TraderRelax IOC limit band from 0.0p to 0.2p
`39=8103=358=NOPLimitExceeded`PB / PoP Intraday Credit Halt0.3 - 0.9 msPrime / Treasury OpsRebalance NOP across AMB Prime / ALB Prime
`39=8103=9958=QuoteStale_117`Co-Located Race / Phantom Pull0.7 - 1.8 msECN Connectivity EngCheck cross-connect jitter vs peer takers
`39=8103=058=PriceTolerance`LP Conditional Last-Look Gate2.0 - 28.0 msLiquidity Mgmt DeskCross-check symmetry & Cover Sheet max hold
Decomposition of 1,840 rejected FIX FX tickets across client collar, NOP credit, race, and last-look causes
Decomposition of 1,840 rejected FIX FX tickets across client collar, NOP credit, race, and last-look causesVerified Exhibit

Mandating Machine-Readable Reject Taxonomy in LP Onboarding#

Regulated dealing venues supervised by the UK FCA, CySEC, and ASIC under algorithmic governance standards (such as MiFID II RTS 6) must maintain deterministic kill-switches and unambiguous error telemetry. Buy-side FX desks refuse to onboard any bilateral FIX feed that returns generic 39=8 | 103=0 | 58=Rejected strings without distinguishing credit, throttle, QuoteID expiration, and last-look tolerance—ensuring complete transparency across quarterly LP scorecard reviews.

Frequently asked questions

Why does FIX Tag 39=4 (Canceled) appear on unfilled Immediate-or-Cancel FX orders?

In standard FIX 4.4/5.0 execution semantics, when a Marketable Limit IOC order arrives after the target price level has already cleared or refreshed beyond the client limit price, compliant matching engines expire the unfilled balance via OrdStatus=4 rather than OrdStatus=8.

How can an execution desk spot an LP hiding last-look rejects behind QuoteStale codes?

If an LP stamps QuoteStale on 20% of directional sweep clips with a 15ms RTT delay while sub-1ms calm clips never collide, latency forensics proves a conditional hold window disguised under a stale-quote text string.

What triggers FIX Tag 103=3 (Order Exceeds Limit) on a Prime-of-Prime session?

Tag 103=3 typically fires at the pre-trade credit gateway when an order would breach either the per-ticket maximum clip size (e.g., $10M cap) or the account aggregate intraday Net Open Position (NOP) margin threshold.

Does GFXC guidance encourage standardized reject reason codes?

Yes. GFXC Last-Look and Cover Sheet working groups strongly recommend unambiguous reject reason codes separating credit, operational/rate limits, and price-validity rejections so clients can audit Principle 17 compliance accurately.

Publication record

Published
14 November 2026
Updated
14 November 2026
Author
Alistair Kerr, CFA
Review
Review not recorded
Automated fact check
Passed in the stored pipeline record
Source URLs
3 recorded

AI assistance recorded: anthropic/claude-opus-4.6. This field does not establish human review or source verification.

Academic & Institutional Attribution

Cite This Research & Embed Live Data Card

Open Attribution License (CC BY 4.0)

Standard academic & institutional research paper citation format.

Alistair Kerr, CFA. (2026, November 14). FIX Tag 39 & Tag 103 Reject Code Forensics in OTC FX Aggregators. FXOpen Ledger. https://fxopenledger.com/orderbook/fix-tag-39-ordstatus-reject-code-forensics-in-fx-aggregators

Regulatory Directory Cross-Reference

Referenced Institutions & Regulatory Registries

Referenced Supervisory Authorities

Explore FX Global Code Last Look Execution Hub (20) →

FXOpen Ledger Data, Benchmarks & Contributor Records

FIX Tag 39 & 103 FX Reject Code Forensics (2026) | FXOpen Ledger