An execution benchmark compares a defined set of trading outcomes with a reference price and a specified treatment of unfinished quantity. Changing the evaluated population can change which execution result appears better.

Original intent and observed fills

A parent order identifies the original intended quantity. Fills record the quantity actually executed at particular prices and times. A filled-price average summarizes those fills; it does not account for quantity that remains unexecuted.

Transaction-cost analysis, or TCA, connects order intent to routing, fills, fees and market context. Its result depends on the decision timestamp, benchmark price and evaluation horizon. Different benchmarks answer different questions about the same activity.

An execution algorithm can trade off completion against price or liquidity-taking costs. The documented IBKR volume-weighted average price, or VWAP, workflow includes a choice that can leave quantity unfinished when avoiding liquidity-taking fees. That is a specific product tradeoff, not evidence that every unfinished order reflects poor execution.

A ranking based on fills alone

Take a buy decision for 1,000 shares at a reference price of 100 dollars per share. Strategy A fills 900 shares at 99.99 dollars and leaves 100 shares unfilled. Strategy B fills all 1,000 shares at 100 dollars. Ignore fees and other costs for this comparison.

A’s filled-price average is one cent lower than B’s. Its executed cost relative to the decision price is 900 × (99.99 − 100), or −9 dollars. B’s executed cost is zero. A looks better under that filled-only price measure.

The comparison has omitted A’s remaining 100 shares. It has not yet evaluated the full original intent.

Adding unfinished quantity under a stated convention

Now set an evaluation price of 100.20 dollars at the chosen horizon. Define unfilled opportunity cost for this buy order as unfilled quantity multiplied by the evaluation price minus the decision price.

A’s unfilled opportunity cost is 100 × (100.20 − 100), or 20 dollars. Adding that amount to its −9 dollars of executed cost gives +11 dollars. B has no unfinished quantity and remains at zero under this constructed measure. B now has the lower measured cost.

The ranking reversal follows from including the omitted quantity under an explicit convention. The 20 dollars is a benchmark valuation of the unexecuted outcome, not a settled cash payment. If the evaluation price or residual convention changes, the result can change.

Selection and comparison boundaries

A filled-only statistic conditions the population on execution. If strategies differ in how much they complete, their averages summarize different portions of the original task. The missing denominator can make favorable executed prices coexist with an unfavorable result for the whole order.

A defensible comparison retains intended quantity, executed quantity, residual quantity, timestamps and the evaluation rule. Comparing real strategies also requires comparable order size, urgency, liquidity and market conditions. This constructed arithmetic does not establish which strategy caused a better outcome in a live market.

When the missing-quantity effect does not apply

If both orders fully complete, there is no unfinished quantity for this mechanism to value. If the question explicitly concerns only fill-price quality, a filled-only average directly answers that narrower question.

The consequence is conditional: omitting unfinished quantity can reverse a ranking when a completion-aware convention assigns a material outcome to that quantity. Neither ranking alone establishes a universal optimal algorithm or satisfaction of a best-execution obligation.

Questions about TCA

Does the lowest average fill price identify the best execution?

No. The original order, completion and benchmark convention determine what the comparison establishes.

Is unfilled opportunity cost a cash charge?

No. It values a missed outcome under a specified benchmark and horizon.

Are filled-only statistics always misleading?

No. They answer a valid question about executed quantity when that population is stated explicitly.