Mechanism and definition first
A “CAD cross” is any forex currency pair where the Canadian dollar (CAD) is one side, but the pair is not quoted against CAD’s usual “home base” framing. In practice, you will often see CAD paired with another currency (for example, CAD vs. a non-USD currency), and the exchange rate expresses how much of one currency equals one unit of the other.
Before using any numbers, define the assessment target:
- Which quote convention you use (e.g., whether the rate is expressed as “1 unit of the base currency equals X quote currency”).
- Which side is CAD (base or quote) for your chosen pair.
- Whether you will use mid, bid, or ask prices for calculations.
This matters because identical underlying market conditions can lead to different computed results if you mix quote direction, base/quote roles, or mid vs. bid/ask inputs.
Data inputs you need
To assess CAD crosses in a way that can be independently checked, collect these inputs.
1) Rate data with explicit quote conventions
You need exchange-rate series or snapshots for the specific CAD cross you are assessing, with clear documentation of:
- Base currency and quote currency.
- Units and decimal precision.
- Whether the provided rate is mid, bid, or ask.
If you intend to compute conversions, also record the exact formula you will use (for example, how you convert through CAD if you are relating two crosses).
2) Price “provenance” (where the data comes from)
Provenance means the origin and method behind the numbers. Record:
- The data provider (platform, aggregator, or exchange feed) and the instrument identifier.
- Whether the data reflects indicative quotes, executable quotes, or aggregated reference pricing.
- Any provider-specific adjustments (for example, if spreads are applied or removed in a way that affects interpretation).
If two sources use different conventions (mid vs. bid/ask; different session windows; different handling of illiquid periods), your comparison can fail even when both are “correct” under their own definitions.
3) Timeliness (when the data is valid)
Timeliness covers the timestamp meaning and the freshness window:
- The time zone and timestamp granularity.
- Whether timestamps represent when the quote was observed, when it was published, or when it became tradable.
- The time window used for any derived calculations.
A common limitation is assuming that because two series are both “live,” they remain aligned. For assessment, document how you will handle non-synchronous timestamps.
4) Transaction cost inputs (for realistic interpretation)
Even without doing a trade, costs affect how you interpret changes in rates. Collect:
- Typical bid/ask spread behavior for the period you analyze.
- Any known commission or fee schedule relevant to your data context.
If you ignore costs, a computed improvement based on mid prices may not be reproducible using executable bid/ask prices.
Quality checks (what to verify before trusting calculations)
Use a checklist that answers “can someone else reproduce my inputs and intermediate steps?”
A) AFVINK points: completeness and alignment
- Are there missing values, and how are they treated?
- Are base/quote roles consistent across the entire dataset?
- Are bid/ask series (if used) aligned in time with the rate series?
B) Evidence or document: quote convention documentation
- Do you have a written description of the pair’s quote format and price type?
- Can you cite the provider’s documentation for how “mid/bid/ask” are defined?
C) Rode flags: mismatches that commonly break assessments
- Mixing mid and bid/ask in the same calculation without stating it.
- Using different timestamp conventions (e.g., different time zones) without converting.
- Combining rate series from sources that use different instruments or identifiers.
D) Klaarcriterium: clear pass/fail for usability
Declare a simple rule: for example, “I only proceed when all required fields (quote type, base/quote, timestamp, spread availability) are present and consistent for the analysis window.”
Evidence or example with stated assumptions
Example of what you should record (without assuming any predictive power):
- Choose one CAD cross and decide whether you use mid or bid/ask. 2) Use snapshots at matching timestamps, or explicitly resample to a common grid. 3) Document the formula you use to compare changes (e. g. , percentage change or difference in quote currency terms) and state your unit assumptions.