Record the conventions that change a quote’s meaning
For each instrument, identify the upstream name, platform name, quote precision, relevant quantity conventions and trading schedule. Record the meaning and time basis of each timestamp. Do not assume that a field named “time” represents the same event at every stage.
Map the feed path from source to receiver and note transformations. A quote can be current at the source yet delayed or rejected at an adapter. Observing only the final terminal view makes that distinction difficult to investigate.
Use a small but representative validation set
| Condition | Check | Evidence |
|---|---|---|
| Active currency pair | Bid/ask, precision, name and timestamp interpretation | Source and receiver samples from the same observation window |
| Market session boundary | Expected open/closed state and quote handling | Session configuration alongside observed behaviour |
| Interrupted source | Freshness alarm and approved stale-data response | Test event, alert destination and recovery record |
| Configuration change | Mapping version and affected receivers | Approved change record and repeated sample checks |
Freshness thresholds belong to the agreed operating requirements. A universal fixed number is not appropriate for every instrument, source or market session.
Make freshness an operational signal
Illustrative feed review: one test pair stops updating while other pairs remain active. A per-connection heartbeat may stay healthy, so the team checks instrument-level freshness and the expected session state before deciding whether the feed has failed.
The handover should contain the mapping register, sample checks, monitoring owners and the response to a missing or stale instrument. Confirm any data-use or redistribution rights with the relevant provider; receiving a stream does not establish those permissions. For a complete integration engagement, use the market-data delivery scope.