Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for accessing a database comprising multiple datasets. Each dataset includes data derived from a respective application. A machine-learning engine determines an analytical rule using at least one dataset of the multiple datasets. The analytical rule is determined by processing input data obtained from the at least one dataset derived from the respective application. A structured dataset is generated based on the determined analytical rule. The structured dataset is generated in response to using the determined analytical rule to analyze data from each dataset of the multiple datasets derived from the respective application. One or more data sequences that represent information flow of a transaction process are determined based on the structured dataset.