I’m currently working with IBM Watson OpenScale and I would like some guidance on how to properly configure the payload and feedback datasets when evaluating a model—especially for Drift v2.

enter image description here

Could you please provide a clear explanation or tutorial covering:

• The expected structure of the payload CSV

• The expected structure of the feedback CSV

• Which columns are mandatory (e.g., predicted label, probabilities, etc.)

• How class probabilities should be formatted (single column vs multiple columns)

• The correct mapping between dataset columns and OpenScale configuration parameters

In my case, I was able to successfully monitor the model by:

• Sending the feedback dataset as the original dataset (including the target column)

• Sending the payload dataset without the target column

enter image description here

enter image description here

This worked for general monitoring. However, when I tried to evaluate Drift v2, I encountered the following error:

“All the values of Class Probability for No column are Nulls” (AQDD1035E)

This makes me think I might be missing something in how I structure or send the class probabilities in the payload dataset.

Could you clarify:

• Whether the payload must explicitly include separate columns for each class probability (e.g., prob_No, prob_Yes)

• If arrays like [0.97, 0.03] are supported or not

• And how exactly OpenScale expects to map probabilities to each class

Any detailed example (CSV format or sample schema) would be extremely helpful.

Thanks in advance!