Preserve order-book captures for exact market replay
Structured dataHead of trading surveillance · Market data engineering
- Decision
- Which market-data captures stay available for replay, and will a replay reproduce the book exactly?
- Why it is hard
- Surveillance and best-execution reviews depend on replaying the market as it was, but captures are large and retention is usually cut first.
- What Saar does
- Compress capture files for retention with exact recovery verified before acceptance.
- Proof boundary
- Replay fidelity also depends on the feed handler. Validate capture completeness and sequence integrity.
Archive FIX session logs without losing the exact wire record
Network payloadsElectronic trading technology · Compliance
- Decision
- Do we have the exact wire-level record for a disputed order or fill?
- Why it is hard
- FIX sessions are logged at volume and often rotated away; reconstructing the wire record later is impossible.
- What Saar does
- Compress session logs losslessly for the retention window, keeping session identity in the index.
- Proof boundary
- Confirm clock and session integrity at capture. Compression does not repair an incomplete log.
Keep tick history reproducible for model research and backtesting
Structured dataQuant research lead · Data platform
- Decision
- Can a backtest be rerun from the exact tick data used the first time?
- Why it is hard
- Tick history is periodically re-cleaned and re-normalised, silently changing what a historical backtest would see.
- What Saar does
- Archive tick extracts as immutable exact objects, separate from the cleaned research copy.
- Proof boundary
- Backtest reproduction also requires the same code and parameters. Validate that separately.
Retain surveillance alerts and supporting data for the full investigation window
Logs & observabilitySurveillance operations · Legal
- Decision
- Can an alert be re-examined with the same data years later, including the messages behind it?
- Why it is hard
- Alert evidence spans feeds, communications and reference data; retaining everything at full fidelity is treated as unaffordable.
- What Saar does
- Archive the alert evidence bundle with exact recovery per component.
- Proof boundary
- Reassessment depends on the analytics version. Compression preserves the evidence, not the model.