Its a session option. when the informatica server performs incremental aggr. it passes new source data through the mapping and uses historical chache data to perform new aggregation caluculations incrementaly. for performance we will use it.

When using incremental aggregation, you apply captured changes in the source to aggregate calculations in a session. If the source changes incrementally and you can capture changes, you can configure the session to process those changes. This allows the Integration Service to update the target incrementally, rather than forcing it to process the entire source and recalculate the same data each time you run the session.

For example, you might have a session using a source that receives new data every day. You can capture those incremental changes because you have added a filter condition to the mapping that removes pre-existing data from the flow of data. You then enable incremental aggregation.

When the session runs with incremental aggregation enabled for the first time on March 1, you use the entire source. This allows the Integration Service to read and store the necessary aggregate data. On March 2, when you run the session again, you filter out all the records except those time-stamped March 2. The Integration Service then processes the new data and updates the target accordingly.

Consider using incremental aggregation in the following circumstances:

You can capture new source data. Use incremental aggregation when you can capture new source data each time you run the session. Use a Stored Procedure or Filter transformation to process new data.

Incremental changes do not significantly change the target. Use incremental aggregation when the changes do not significantly change the target. If processing the incrementally changed source alters more than half the existing target, the session may not benefit from using incremental aggregation. In this case, drop the table and recreate the target with complete source data.

Note: Do not use incremental aggregation if the mapping contains percentile or median functions. The Integration Service uses system memory to process these functions in addition to the cache memory you configure in the session properties. As a result, the Integration Service does not store incremental aggregation values for percentile and median functions in disk caches.