A weed emergence prediction model, WeedCast, was used as a decision aid to schedule potato cultivation with and without herbicides at Wooster, OH, USA; Charlottetown, PE, Canada; and Fargo, ND, USA, from 2001 to 2003. Studies were laid out in a split-plot design with herbicides (±) forming the main plots and cultivation timing as subplots. Cultivation was done at 15, 30, or 60% of predicted weed emergence. Subplots were either left unsprayed or treated with metolachlor + metribuzin at 1.68 + 0.5 kg ai ha−1 and only cultivated at predetermined timing. Cultivation timing was based on predicted emergence of common lambsquarters at Wooster and Charlottetown, whereas eastern black nightshade was the indicator weed at Fargo. Weed control for the different cultivation timings varied among sites and years and was consistently better in plots where herbicides were followed by cultivation. Cultivation alone resulted in poor weed control and significantly reduced potato tuber yield compared with those in plots where weed control also included herbicides. Use of herbicides followed by cultivation and hilling increased tuber yield by 4.6, 4.3, and 8.7 t ha−1, when cultivations were done at 15, 30, and 60% of predicted weed emergence, respectively, and 12.2 t ha−1 for hilled-only plots. The average potato yield increase at Charlottetown was 9.7, 5.9, 6.9, and 7.4 t ha−1 for hilled-only plots and for hilled after cultivations at 15, 30, and 60% of predicted weed emergence with herbicides, respectively. There was no apparent pattern for treatment effects at Fargo, and the potato tuber yields were greatly reduced mainly because of excessive precipitation during potato establishment. Use of WeedCast as a decision-aid tool could be an asset in determining when to do the first and subsequent cultivations. It may work best for growers who use cultivations in potato to remove weeds that were not controlled by herbicides.