Due to heavy computational burdens, state-of-the-art demand-side bidding models have to sacrifice the accuracy of uncertainty estimates in exchange for tractability, and therefore fail to derive a high bidding revenue as expected. This paper analyzes the entire bidding process, from scenario reduction to bidding curve construction, to find out the weak link that needs improvement. With this purpose, we develop a novel process to save some computational resources by decomposing the complex bidding model, then reallocates these resources to consider more scenarios and more precise bidding curves. This process is fully decoupled and is able to achieve both a bidding revenue growth and a significant calculation speedup. In order to reduce computational complexity, the scenario reduction and bidding steps are separated, reorganized and decoupled. The price response zone and single-period scenario reduction are specially designed for temporal decoupling. As for the bidding curve construction, a fine-grained bidding model is carefully developed to meet the high-precision requirement. The proposed process and models are demonstrated through a typical end-user who constructs piecewise linear bidding curves in a day-ahead market. Extensive case studies verify the fast computing speed and bidding revenue growth of the proposed methodology.