Most markets have a U-shaped intraday volume curve: heavy at the open, thin at midday, heavy into the close. A VWAP algo front-loads and back-loads the schedule to match, trading more shares when more liquidity exists.
It is measured against the day's actual vwap, which makes it popular as a benchmark for agency execution: beating VWAP is a defensible definition of a good job. That benchmark is also gameable, since an algo can guarantee near-VWAP by simply being average.
The forecast is the weak point. Volume profiles shift on index rebalance days, expiry days, and around news, and an algo following yesterday's profile into an unusual day will be badly scheduled. Most implementations blend a historical profile with live volume tracking.
Related: twap-algorithm, vwap, execution-algorithm, pov-algorithm