2026-09-25 - Event Data and Change Data Handled Distinctly in History¶
Release¶
AgileData.io - Right Pattern, Right Data: Events and Changes Loaded the Way They Should Be
Data that records things happening, such as transactions, page views or video plays, behaves differently from data that describes the current state of something, such as a customer or a product. Our history loading patterns have been refined to treat these two kinds of data distinctly from landing all the way through to detail.
For event data, each event is now loaded the first time it is seen, and re-delivered copies are ignored, even when they arrive on a later day. The grain you define on the Input Table, meaning the columns that identify one event, is used to decide what counts as the same event. For change data, the effective date stays part of the comparison, so changes within a day, and changes back to an earlier value, are captured.
The guidance our AI agents use for these patterns has been updated to match. That includes the advice to always specify the business date an event happened, so downstream tiles report when the event occurred rather than when the file arrived.
What’s New:
Event data history loads ignore re-delivered rows across days
Input Table grain used to identify unique events
Event data supported as a source pattern for all event outputs
Agent guidance updated for event and change data loading
What this means for you:
No duplicate events when source systems re-send overlapping data
Accurate change history for records that update over time
The groundwork for the Load Type override released in October
Last Refreshed¶
Doc Refreshed: 2026-10-21