A note on end-of-day runs past its window
Getting UDSRs and the end-of-day result calculator right
User-defined simulation results are where Endur valuation logic lives — and where end-of-day quietly goes slow or wrong. Patterns for UDSRs that stay fast and correct.
A missing valuation that reads as 0.00 doesn't trip any alarm and flows straight into P&L.
If you want to know what an Endur platform actually values and how, read its UDSRs. User-defined simulation results are where the bespoke valuation and reporting logic lives, and they’re disproportionately responsible for both the numbers the desk trades on and the length of the end-of-day run. They reward being written carefully and punish being written quickly.
A UDSR is a contract, not a script
It’s tempting to treat a UDSR as “some JVS that produces a column.” It’s more useful to treat it as a contract: given a defined set of simulation inputs, it returns a result of a known shape, for every deal it’s asked about, every time. Holding that line — same inputs, same output, no hidden dependence on run order or external state — is what makes results reproducible and end-of-day debuggable. A UDSR that quietly depends on something outside its inputs is the one that gives a different number on the overnight run than it did when you tested it at lunchtime.
Compute once, in the simulation
The most common performance mistake is doing per-deal work that should have been done once. The OpenLink library gives you the simulation result set and the pricing already produced upstream; a good UDSR reaches into what’s already there rather than re-deriving it deal by deal. Re-pricing inside a result calculator, re-querying reference data per transaction, or rebuilding a curve you were already handed — each is fine on a test portfolio and quietly fatal across the full book at end-of-day.
The discipline: pull what you need from the simulation once, hold it, and let the per-deal step be a lookup rather than a calculation.
Fail loudly, never silently
When a UDSR hits a deal it can’t handle, the worst outcome is a plausible-looking zero. A missing valuation that reads as 0.00 doesn’t trip any alarm and flows straight into P&L. Decide explicitly what “I cannot value this” looks like — an error result, a flagged status, an exclusion you can see — so that a gap announces itself rather than hiding inside an aggregate. End-of-day reconciliation is only meaningful if failures are visible.
End-of-day is the real test
A UDSR that’s correct and quick on a handful of deals tells you very little. The honest test is the whole book, on representative data, timed. End-of-day is a long chain of these calculators feeding results into the next stage, and a single slow or wrong link shows up as either a run that overshoots its window or a number nobody can explain in the morning. Test against the full population, reconcile the output against a known-good baseline, and watch the timings — a calculator that doubles in cost when the book grows is a problem you want to find in test, not in production at 2am.
In short
Treat each UDSR as a small, well-defined, well-behaved function: clear inputs, predictable output, work done once, failures made visible. Do that and the result calculator becomes the dependable core of end-of-day rather than its most common source of late-night surprises.
Is this happening on your desk?This is the consulting work I do. Tell me what you're seeing and I'll say whether I can help.
Send a consulting enquiry