How Does Canals Learn?

Learning Occurs Only from Submitted Data
Learning in Canals happens exclusively based on the data that is submitted from Canals to your ERP. If a user selects products for a job but never submits that job to the ERP, those selections will not contribute to the learning process. Only submitted jobs are used to train the model.

Learning Happens Overnight
Once a job is successfully submitted from Canals to your ERP, learning takes place overnight. If a user submits a job and then resends the exact same job into Canals shortly after, there typically won't be learning results right away. This is because the system needs time (usually overnight) to process the data. However, in some cases, we may perform intraday learning, but it's not guaranteed.

Resending Jobs and Learning Results
If a job is resubmitted the next day, users should expect to see learning results. In fact, many products may already be auto-selected based on the learned data. However, if a previous job for the same customer had conflicting product selections for similar requests, both options will likely rank highly in the results, but auto-selection won't occur until the model has had more time to resolve the conflicting data and discern the variables at play.

Factors Affecting the Speed of Learning
The speed and effectiveness of the learning process depend on several key factors:

• Volume: The more data there is, the faster the model can learn.

• Breadth: A wider range of scenarios and product types leads to more robust learning.

• Clarity: Clear connections between a customer’s request and the chosen product (e.g., matching dimensions) speed up the learning process.

• Consistency: Consistent patterns in the data help the model learn quickly. If there’s a lot of inconsistency in the choices, the model may take longer to understand the relationship.