A database trigger is the most efficient design as it operates directly at the data source level, independent of the Appian execution engines. When a record is modified via any Appian action, the database trigger automatically captures the changes and writes them to an audit table within the same database transaction. This approach completely offloads the auditing logic from the Appian application, resulting in the lowest possible performance impact on Appian's servers. The auditing process is transparent to the Appian objects (e.g., process models, interfaces) that perform the data modification, ensuring minimal overhead and complexity within the application itself.

1. Appian Corporation. (2023). Data-Centric Design. Appian Documentation. In the "Leverage the Database" section

the documentation advises

"Leverage the database for what it does best: managing and manipulating large sets of data... For example

if you need to update thousands of rows in a database table

it is much more efficient to call a stored procedure from Appian than it is to loop through those updates in a process model." This principle directly supports offloading the row-level audit task to a database trigger for maximum efficiency.

2. Appian Corporation. (2023). Record Events. Appian Documentation. This document describes the Appian-native approach of using record events to trigger a process. While a valid pattern

this method still consumes Appian application resources to handle the event and execute a process

making it less efficient and having a greater impact than a database trigger that operates outside the Appian execution engine.

3. Silberschatz

A.

Korth

H. F.

& Sudarshan

S. (2019). Database System Concepts (7th ed.). McGraw-Hill. Chapter 8

"Advanced SQL

" discusses the use of triggers for maintaining data integrity and creating audit trails. The text explains that triggers are executed by the database system automatically as a part of the modifying transaction

which is inherently more efficient for auditing than an external application-layer process.