Electric load forecasting has been integrated into the business decision making processes across virtually every segment of the power industry. Power companies use short-term and long-term load forecasts primarily for power systems operations and planning, while electricity retailers use load forecasts for pricing and procurement decisions. In power delivery systems, a delivery point is a node where the electricity is delivered to the distribution network in order to supply power for a local area. Load forecasts at the delivery point level provide values for distribution system operators. Load forecasters face two major challenges when forecasting the load profiles at the delivery point level: data quality and randomness of the load. Data quality issues play a vital role in producing accurate forecasts. In the power system hierarchy, the frequency of events causing the quality issues for load data is unique and more intense at the delivery point level. In this research, a delivery point load forecasting framework is proposed, which includes different components focusing on the quality issues in both load data and weather data, such as load transfer detection; meter grouping; load anomaly detection; and weather data cleansing using multiple load zones. The framework is developed and evaluated using the data from a distribution company in the United States. Enhancements of data quality in each step are evaluated within a load forecasting process. The effectiveness of the proposed solution has been empirically confirmed through significant improvements to the forecast accuracy.