Current forecasting methods are generally manual and based on historical data only. Typical forecasting methods involve manually identifying the same week in a previous year or year(s), which are considered to be similar to the current year, and manually adjusting based on experience and gut feel to account for new information like competitive activity, price movements, promotions and weather. Simple statistical forecasting often doesn’t provide an improvement over manual methods based on experience. However, machine learning based methods allow more variables such as broader price data, weather data, public holidays, geographical and socioeconomic data to be incorporated leading to greater accuracy than manual methods.