Forecasting accuracy is an important issue for effective decision making in the areas of strategic planning, production planning, inventory management and other areas. In particular, this aspect is most relevant for items that have demand patterns with important levels of intermittency and lumpiness. This work describes the efforts of a Mexican company to look for opportunities to reduce a high level of customer service variability through a better selection of forecasting procedures and the application of demand aggregation to all the items of the company´s catalogue. Initial results are obtained from a pilot study carried out in all the stores of the Puerto Vallarta plaza of the company. The resulting forecasting mean squared error (MSE) was decreased significantly in the range of 35 to 60%. The firm estimates a reduction in the order of 21% of the variability in service level (standard deviation) due to the improvement in forecasting accuracy.