There are certain requirements for any algorithm to be able to produce accurate predictions over time and to be trained and learn. Please be aware of these before using any Max AI features (i.e. AI segments).
Emarsys collects data from various sources such as channel engagement, web behavior tracking information, and uploaded order data to your account. The following data tables are populated and required to run AI Segmentation:
- Email sends
- Email clicks
- Email opens
- Web Extend sessions
- Smart Insight sales
- Smart Insight contacts
As additional criteria, you need to have:
- All of the above tables simultaneously (if any of them is missing, some or all AI segment calculations will fail).
- At least 7 months of respective data collected.
- The last 6 months of backtests should be successful.
- Model performance must be above a certain threshold.
Please note that:
- If you initiate a reload of your sales data, until the data is completely reloaded the platform cannot decide whether or not your uploaded data is eligible to run AI predictions. Eligibility calculations are halted for a week to leave sufficient time to finalize the reload.
- Performance of AI predictions could highly vary for different periods.
- Usually predicting engagement is easier than predicting purchases, and predicting purchases for leads is the most difficult one since they do not have purchase history.
- Having the data available does not guarantee successful predictions. We perform tests on the data to make sure it is accurate enough to train our models.
If you believe you meet the above criteria, please contact your Customer Success Manager to validate readiness to run AI Segmentation.