BDS / Intraday usability
Delayed Alerts and Missed Trades: Keep the Difference Visible
Delay, absence and an unfilled order are different reasons an instruction may not become a subscriber trade.
Keep separate clock intervals
Source publication to receipt concerns delivery. Receipt to review concerns attention. Review to order acceptance concerns the account workflow. A single delay figure can conceal which part was observed and which was assumed.
Use absolute instants and the provenance of each timestamp. If one time is unknown, do not calculate a precise latency from the remaining ones. A retrospective message label is not necessarily a contemporaneous receipt log.
Distinguish missed states
An unavailable reader, expired instruction, incompatible instrument and unfilled order are different states. Retain their reasons instead of assigning them a common assumed source entry. A service can have prompt delivery and still be unusable during the reader's ordinary schedule.
Keep repeated notifications linked to the first instruction. They may help retrieval without constituting fresh eligible setups.
Worked price-delay case
Hypothetical long setup: reference entry 50.00, stop 49.75 and target 50.50. A later assumed entry at 50.10 changes the planned distance to the same stop and target. This is arithmetic under fixed levels, not an instruction to chase the price.
The original validity rule still comes first. If it says the entry is no longer eligible, a plausible later price does not rescue the setup. The workbench shows the difference without inventing an actual subscriber fill.
Avoid the best delayed reconstruction
A replay should preserve its frozen research observation-window rule. A subscriber study should use the actual receipt and account records. Do not take whichever price looks best from either process and call it one verified result.
An internal research fill convention can be valid for its labelled purpose without representing attainable customer execution. The publication does not modify or certify that model convention.
Compare a complete session sample
Include ordinary busy intervals and missed observations. A fastest-message anecdote is not an end-to-end latency benchmark. A benchmark needs a declared environment, clocks, source events and genuine logs.
A small personal log can reveal a mismatch, but cannot establish that one provider is always faster or profitable after delay. Keep the conclusion proportional to the observed evidence.
End with an operating limit
State which delays or unavailable windows prevent the intended workflow. Request the original validity and retrieval policy where needed. The choice can be to defer, observe without positions or use an existing tool.
We have not purchased feeds or measured provider latency. All numerical timing examples in this chapter are hypothetical and should remain labelled when quoted.
Missed is not a favorable hypothetical fill
An instruction can be part of the full publication population without entering the subscriber trade set. Keep that distinction. Filling a missed row with a reconstructed attractive price hides the usability finding and changes the execution claim. A separate labelled research replay can exist without becoming the customer experience.
Preserve clock failure cases
If receipt appears before publication, inspect offset, timestamp meaning and source provenance before calling the delay negative. The calculator rejects impossible entered chronology rather than reporting an advantage. Genuine clock uncertainty needs a narrative or interval, not an exact arithmetic result built from incompatible values.
What a real benchmark would need
A genuine test needs frozen source objects, comparable clocks, routing environment, event logs and ordinary/busy observation windows. It should retain failures and uncertainty. None of these are created merely by publishing a calculator. BDS supplies the proposed fields and hypothetical arithmetic, not an invented benchmark result.
Evidence Packet
| Evidence object | Record to retain |
|---|---|
| Delivery | Publication and receipt with provenance |
| Attention | First review and ordinary availability |
| Execution | Actual order state or explicitly hypothetical price |
Related Session Questions
Primary References
The sources support documented mechanisms and source definitions. Our procedure and teaching examples are not completed paid-feed, latency or account tests.