A new Mule application under development must implement extensive data transformation logic. Some of the data transformation functionality is already available as external transformation services that are mature and widely used across the organization; the rest is highly specific to the new Mule application.
The organization follows a rigorous testing approach, where every service and application must be extensively acceptance tested before it is allowed to go into production.
What is the best way to implement the data transformation logic for this new Mule application while minimizing the overall testing effort?