Appian stores every process-variable value in the process engine.

Memory can therefore be reduced by 1) passing needed data directly to an activity or smart-service input (A), so it is never written to the engine; 2) declaring process variables that are only needed while the node executes as activity-class parameters (D), which keeps the data in the activity’s JVM heap and not in the engine; and 3) avoiding the storage of large strings, documents, or lists in process variables (F), thereby cutting the biggest single contributor to the engine’s memory footprint. These three techniques are explicitly called out in Appian’s performance-tuning guidance and have a measurable impact on runtime memory consumption.

B. Sub-processes are useful for modularity; memory impact comes only from what you pass, not their mere existence.

C. The count of variables adds negligible overhead; size of stored values is what matters.

E. Breaking a node into several nodes affects execution logic, not how much data is persisted in memory.

1. Appian Product Documentation 23.2 – “Performance Considerations > Process Memory”

bullets “Feed Data Directly to Smart Services” and “Use Activity-Class Parameters”.

2. Appian Product Documentation 23.2 – same section

paragraph “Avoid Storing Large Data Sets or Large Blocks of Text in Process Variables”.

3. Appian Academy Online Course “Process Modeling Fundamentals”

Lesson “Improving Process Performance”

slide set “Reducing Process Memory”

slides 3-6 (accessed 2025-09-29).