5.6 Execution benchmarks: VWAP, TWAP and implementation shortfall

Motivation

"Did the desk execute well?" needs a benchmark. The choice of benchmark changes behaviour, so it is a design decision, not an accounting detail.

5.6.1Benchmark definitions#

Definitions

For a set of trades with prices $p_i$ and quantities $v_i$ over a window:

\text{VWAP} = \frac{\sum_i p_i v_i}{\sum_i v_i}. \tag{5.14}

TWAP is the time-weighted average price over the window; a TWAP schedule splits an order into $N$ equal child orders at equal time intervals. A VWAP schedule splits it in proportion to expected volume in each interval, $q_k = Q\,\hat v_k/\sum_j \hat v_j$ , which with a U-shaped volume profile means trading more at the open and close.

Implementation shortfall (Perold 1988, "The implementation shortfall: paper versus reality", Journal of Portfolio Management 14(3)) compares a real execution with a paper portfolio that traded the whole order at the decision price $p_d$ (the mid when the decision was made). For a buy of intended size $Q$ of which $Q_e$ executed at average price $\bar p$ , with the price at the end of the horizon $p_T$ and the price at the start of execution $p_0$ :

\text{IS} = \underbrace{Q_e(p_0 - p_d)}_{\text{delay}} + \underbrace{Q_e(\bar p - p_0)}_{\text{execution}} + \underbrace{(Q - Q_e)(p_T - p_d)}_{\text{opportunity}} + \text{fees}. \tag{5.15}

(For a sell, reverse the signs of the price differences.)

5.6.2Achieving the VWAP#

Proposition 5.4a volume-proportional schedule achieves VWAP

Divide the window into intervals $k = 1..K$ with market volume $V_k$ and market VWAP $\bar p_k$ in each. If a trader executes $q_k = Q V_k/\sum_j V_j$ in interval $k$ at the interval VWAP $\bar p_k$ , her average price equals the window VWAP.

Proof

Her average price is $\sum_k q_k\bar p_k/Q = \sum_k V_k\bar p_k/\sum_j V_j$ . Since $\bar p_k = \sum_{i \in k} p_i v_i / V_k$ , this is $\sum_k\sum_{i\in k} p_i v_i/\sum_j V_j$ , the window VWAP.∎

The difficulty in practice is that $V_k$ is unknown in advance (it must be forecast), and that the trader's own volume is part of $V_k$ and moves $\bar p_k$ .

5.6.3VWAP as benchmark and settlement price#

Why VWAP is the standard benchmark Berkowitz, Logue and Noser ("The total cost of transactions on the NYSE", Journal of Finance 43(1), 1988) proposed measuring execution cost against the day's VWAP. Its appeal: (i) it is the average price paid by all volume, so beating it means doing better than the average participant; (ii) it is computable from public data; (iii) for an order that is a small share of volume it is hard for the trader to manipulate. Its weaknesses: (i) a trader who is a large fraction of volume pulls VWAP towards her own price, so she "beats" a benchmark she created; (ii) it ignores delay and opportunity cost, so a desk paid to match VWAP has no incentive to complete urgent orders quickly; implementation shortfall fixes this.

VWAP as a settlement price The same robustness makes VWAP a good settlement price. Before August 2026, NSE's official closing price was the VWAP of the last 30 minutes; from 7 Dec 2026 expiring derivatives again settle on the VWAP of the last 30 minutes of continuous trading rather than on the closing auction price (SEBI circular of 15 Oct 2026, as summarised; M01 · 1.2). To move a 30-minute VWAP by a given amount, a manipulator must move the price for a substantial share of 30 minutes' volume, which by (5.13) is expensive; moving a single last trade is cheap.

Practitioner notes

Execution algorithms offered by Indian brokers to institutions (VWAP, TWAP, POV "percentage of volume", IS) implement these schedules. A retail trader cannot use them, but their footprints (heavy volume at open and close, child orders at regular intervals) are part of the intraday tape.
Benchmarks are incentives. Choose the one that matches the investment decision: IS for alpha-driven trades, VWAP for orders with no urgency, close for index rebalances.