In a distribution channel, product manufacturers will reward retail traders who achieve a high value of sales. In order to obtain reward from the manufacturer, distributors might form an alliance where a cheating retail trader could accumulate the amount of sales via collaborative fraud with other distributors. In order to detect such fraud behavior, we mainly focus on answering three questions: WHO conduct the sale accumulation? WHICH sale records are modified? HOW they conduct the sale accumulation? However, if any one of the three questions cannot be answered, the other two may not be handled either. To cope with this difficulty, we propose the new concepts of data chunk and partially ordered lattice of the data warehouse schema. Based on these concepts, we propose the concepts of sale accumulation point(SAP) and sale accumulation pattern(SAPAT). SAP can answer WHICH sale record is accumulated by WHO. SAPAT can answer HOW the distributors conduct the sale accumulation. By designing algorithm to detect SAP and SAPAT all together, the three questions can be answered at the same time. The experimental results demonstrate that the proposed algorithm based on the SAP to solve the WHICH and WHO problems can achieve an average AUC value of 0.65. Traditional feature extraction methods can only achieve an average AUC of 0.25. The proposed algorithm based on the SAPAT to handle the HOW problem can achieve an average accuracy of 75.2%. The baseline algorithm can only achieve average accuracy of 27.1%. Furthermore, the experiments demonstrate that the proposed method can well handle the WHICH and HOW problems on both synthetic and real data.