Redesigning Trade Fails Management for T+1 and Beyond
Settlement fails have always been an accepted cost of doing business. For decades firms managed them the same way: wait for the depository rejection or the custodian status message, print the fails report overnight, and set the settlements team to work the queue the next morning. That model survived because the market gave it room. A two-day cycle left a full buffer day between trade capture and settlement, and a fail, once discovered, could usually be repaired within a day or two without anyone outside operations noticing.
That room is gone. T+1 removed the buffer day in North America, CSDR settlement discipline attached a daily cash penalty to every fail in Europe, and Treasury clearing and the 2027 European migrations tighten the vise further. Meanwhile the information needed to see a fail coming, unmatched confirmations, missing settlement instructions, projected inventory shortfalls and funding gaps, already sits inside firms' own systems on trade date. The unit of work is shifting from the failed trade to the at-risk trade.
Automating a reactive process produces a faster reactive process. The redesign is a different exercise: predict likely fails before intended settlement date, route them intraday to the teams that can act, automate the remediation that does not need a human, and measure the hardest thing an operations team produces, which is the fail that never happened. The full report sets out the scoring model, the policy line between automated and human remediation, the control evidence a scoring model owes under SR 11-7 and SS1/23, and a defensible Prevented-Fail Rate.
Selected Conclusions
• The unit of work is shifting. Fails management is being redesigned around the at-risk trade rather than the failed trade, because T+1 and settlement discipline have made intercepting the signal cheaper than repairing the fail.
• The fail is visible before it happens. Most fails telegraph themselves through signals already resident in firm systems, unaffirmed or unmatched status, missing or stale settlement instructions, projected inventory shortfalls and funding gaps, and demonstrated prediction capabilities in the Treasury market confirm the signals carry real information.
• Prevention can be measured honestly. A credible prevented-fail metric compares model-expected fails with observed fails across intervened trades, validated against non-intervened trades, which gives the function a production number rather than an absence of bad news.
