A method for detecting granular data store changes in a computing environment. Data values stored in a first relational database are changed. Each data value is stored in a row and a column of the first relational database. Rows (i.e., changed rows) including the changed data values are identified. Changed rows are propagated to a second relational database. Columns including the changed data values are identified in the second relational database. The types of changes associated with the changed data values are identified. Filters optionally filter out one or more of the changed data values from further processing. At least one data value of the changed data values, along with their types of changes, are transformed into a language neutral format included in packaged data to be sent in response to a client requesting the at least one data value.