Within the Transaction Manager "Configurations" tab, you'll find the "AutoPush Settings" section.
Within the Configurations tab, the Automation Settings section is where you will map a Smart Connect account to a GL account from the QuickBooks Online (QBO) or Xero Chart of Accounts. The start date will be when transactions should start “AutoPushing” to the GL. The AutoPush start date can be set to cover transactions up to 24 months in the past.
The Additional Automation Settings section is where you can configure additional settings, such as: ML Training Start Date, Materiality Threshold, Default Assignee, and Default Settings for Low/Medium Confidence Transactions.
-
The “ML Training Dates" Start Date allows you to set a specific date when the ML models should start training based on the historical GL data. If no date is defined, the models will look at all GL activity. This is useful in case there’s historical data we do not want the model to train from.
-
Users may choose both a start and end date to create a specific range of data to train machine learning.
- When the start or end dates for the machine learning models are changed, the models will retrain immediately. Predictions based on the retrained model will begin within 24 hours.
- The Uncategorized Accounts setting allows you to exclude accounts from machine learning to help improve the accuracy of your future automated transaction categorizations.
-
The Materiality Threshold allows you to set an amount where all transactions greater than or equal to a certain amount will automatically post to the Needs Review tab for further review.
-
The Default Assignee Settings allows you to select a user that should be assigned to all transactions on the Needs Review tab.
-
The Settings for Low/Medium Confidence Transactions allows you to select a default category and payee to which all low and medium confidence transactions should be posted. For example, you can post all of these transactions to “Uncategorized Expense” and “Uncategorized Vendor.” If these settings are not configured, the transactions will post to the category and payee that the ML model predicted.
Filtering transactions
AutoPush Confidence Level Filter:
On both the Needs Review tab and “Processed” tabs, you can filter transactions based on their AutoPush confidence levels:
- High 98% and greater
- Medium 90% through 97.99%
- Low 0% through 89.99%
Transaction State Filter:
You can also filter them based on whether they are Bank Feed or General Ledger on those same tabs.