When you publish transformations for object-centric process mining, we create complete transformations, using your scripts as the core of them, to populate the tables in the Celonis database of objects and events. Previously, we set the data type for any attribute column containing a string to VARCHAR(255). Now, we can set the string length for each attribute column according to what’s in the source data that the transformation is handling - either the specified length for the column, or the database default length for VARCHAR columns. If you set a VARCHAR length in your original transformation script, for example using the CAST or LTRIM functions, we'll continue to use that.

Optimizing the VARCHAR length means your transformations won’t fail due to overlength strings, and optimizes the performance of the Celonis database, which is very responsive to changes in string length. However, you should be aware that if the string columns in your source database are always or often set to a length greater than 255 characters, the performance of the Celonis database could potentially deteriorate rather than improve. This feature is currently in Limited Availability and will be released to all teams over the next few weeks. You’ll see the changes becoming effective after publishing in objects and events to change the generated transformations.