Choosing an Execution Benchmark

A portfolio manager decides to buy 200,000 shares when the stock is at $40.00. The order reaches the trading desk a few minutes later at $40.02. The desk works it through the day, averaging $40.09 a share. The stock closes at $40.25, and the volume-weighted average price over the working interval was $40.11.

Now grade the trader.

Benchmark	Level	Slippage on the buy
Decision price	40.00	+22.5 bps (cost)
Arrival price	40.02	+17.5 bps (cost)
Interval VWAP	40.11	−5.0 bps (beat it)
Closing price	40.25	−39.8 bps (beat it)

One execution, four verdicts, spanning sixty basis points. Nothing about the trade changed. Only the question did.

What each benchmark is actually asking

• Decision price — "what did the whole idea cost, end to end?" It includes the delay between the manager's decision and the order reaching the desk, which the trader usually does not control. Honest for the fund, unfair as a trader scorecard.
• Arrival price — "from the moment you owned this order, how much did you give away?" This is Implementation Shortfall as traders are normally measured on it. It charges you for both market impact and the price drifting away while you were patient.
• Interval VWAP — "did you beat the average participant?" It forgives drift entirely: if the stock ran 25 cents, VWAP ran with it.
• Closing price — "did you get the mark?" The right question when your fund is valued at the close, or when the trade exists to track an index that rebalances there.
• TWAP and PWP-n (the price of the first $n \times$ your size traded after arrival) are variations on "average", used where volume data is thin or manipulable.

Why VWAP flatters you: the self-reference problem

Interval VWAP includes your own trades. Here the interval printed 500,000 shares: your 200,000 at $40.09 and everyone else's 300,000 at $40.1233, giving

$\text{VWAP} = \frac{200{,}000 \times 40.09 + 300{,}000 \times 40.1233}{500{,}000} = 40.11.$

Now suppose the desk had traded badly and averaged $40.20 instead — 11 cents worse, $22,000 of real money. The VWAP itself moves to $(200{,}000 \times 40.20 + 300{,}000 \times 40.1233)/500{,}000 = 40.154$, so the measured slippage goes from −5.0 bps to only +11.5 bps. A 27.5 bps deterioration in reality shows up as 16.5 bps on the report.

That damping factor is exactly $1 - \text{participation rate} = 1 - 0.40 = 0.60$. At 40% of the volume you can only ever be graded on 60% of your own mistakes; at 80% participation, VWAP is almost a mirror.

Matching the benchmark to the reason for the trade

Why you are trading	Use	Because
Fast-decaying alpha signal	Arrival price	Delay is a real cost and must be charged
Index tracking, fund flows	Closing price	Your NAV is struck there
Patient rebalance, no timing view	Interval VWAP / TWAP	Being average is genuinely the goal
Broker risk transfer / guaranteed VWAP	Arrival, with the guarantee priced in	Risk was transferred, so price it up front
Illiquid names, thin prints	TWAP or PWP	Volume-based benchmarks are too easy to distort

In interviews

The reliable question is "you beat VWAP by 5 bps — good day?" The expected answer is it depends what the trade was for, followed by the self-reference point: if you were 40% of the interval's volume, beating VWAP by 5 bps is close to beating yourself. Then contrast it with arrival price, where the same execution cost 17.5 bps, and say plainly which one you would put on a scorecard and why. Being able to name the incentive each benchmark creates — VWAP makes you patient, arrival makes you aggressive — is the part that shows you have thought about it as a control system, not a report. See TWAP, VWAP & POV for the algos and Implementation Shortfall for the full cost decomposition.