When you read the hierarchical data in a Source transformation before passing it to the Hierarchy Processor transformation, use JSON format instead of an intelligent structure model whenever possible. However, use an intelligent structure model for files that are larger than 1 MB.

Avoid non-splittable files that are larger than 100 MB. Instead, process several small JSON files.

Reduce the number of nested hierarchies. Mappings process a three-level hierarchy faster than a five-level hierarchy.