Direct answer
Execution quality in No Dealing Desk (NDD) contexts is best assessed as a set of observable process outcomes: how quickly orders reach execution, how fills relate to the quotes available at decision time, and how costs and price adjustments affect the final result. Because real markets change and providers may differ, you should treat performance metrics as conditional and verify them using consistent assumptions rather than expecting any single measure to guarantee good results.
Mechanism or definition
“No Dealing Desk” generally refers to an execution approach where the provider does not (by design description) act as the counterparty in a traditional dealing-decision loop for each client order. Even without committing to a specific provider’s internal design, you can still evaluate execution quality by focusing on externally visible behaviors:
- Order-to-execution speed (latency): the time between sending an order and receiving an execution confirmation.
- Fill characteristics: whether fills cluster around the contemporaneous available price, and whether partial fills occur frequently.
- Price handling during execution: how the system manages changes between when a trader sees a price and when the order is executed (for example, during brief liquidity gaps).
- Cost transparency impact: how spreads, commission-like fees, and any execution-related markups combine into total transaction cost.
When you assess these, separate stable mechanics (the general execution process and matching behavior you can test repeatedly) from variable conditions (market volatility, liquidity, and slippage risk).
Evidence or example
A practical way to assess execution quality without relying on promises is to run a small, consistent set of tests and compare outcomes under the same assumptions:
- Define the measurement window. Choose specific times and a short list of instrument conditions (e.g., “normal liquidity hours” versus “high-volatility windows”).
- Record inputs at decision time. For each order, capture the displayed or referenced price at the moment you submit, and the timestamp when you submit.
- Record execution outcomes. Capture the fill price(s), timestamps of fills, and whether the order was filled fully or partially.
- Compute differences using explicit assumptions. For each order, calculate price deviation = (average fill price − reference price at decision time). Track the distribution, not just averages.
- Separate costs from execution. Repeat comparisons while accounting for total cost (spread plus other stated charges). If costs differ, ensure you compare on net effective price rather than only the quote-side number.
Material example of a limitation in the method: if you only look at average deviation across all conditions, you can hide a failure mode where executions are acceptable during calm periods but degrade sharply during liquidity gaps. A distribution (for example, median and tail outcomes) is more informative.
Limitations and risks
Several limitations can make “execution quality” appear better or worse than it truly is:
- Market condition confounding: Higher volatility and thinner liquidity raise the chance of unfavorable deviations, regardless of the provider’s design.
- Reference-price mismatch: The price you observe at submission may not be the exact tradable price available at execution time, especially during fast price moves.
- Partial fill complexity: Partial fills can make “best execution” look worse if you only compare the first fill or ignore total net cost.
- Latency and queue effects: Even if an order is routed without dealer intermediation, network delays and matching/queueing can cause executions at worse prices.
- Non-repeatability: Historical relationships do not establish future results; execution behavior can change when market structure changes or when internal systems update.
A material failure mode to explicitly look for is unfavorable tail outcomes: most executions may be close to reference prices, but rare events during liquidity gaps or rapid moves can produce outsized deviations. This is why you should test across multiple conditions and focus on outcome distributions.
Verification or next question
To verify your assessment independently, rely on records you can capture: timestamps, reference quotes at decision time, fill prices, and whether fills were partial. Then check whether your conclusion holds across different market regimes (calm vs volatile) and across repeated test runs. If your results depend heavily on one narrow window, treat the conclusion as incomplete.