This paper proposes a framework for the extract-transform-load (ETL) of the Chinese style cross-tab. It makes a reverse engineering process for the Chinese style report. That is to say, converting the report to its original scheme of relational database table, and then making an ETL processing on the two-dimensional table. And ultimately load the data into the fact tables. Finally, we discuss implementation issues and we present the design and experiment of the system based on our framework.