The probability column will usually result as null in cases where an x-variable used in the model is expected as a particular variable type, but the dataset being scored stores that same x-variable as another variable type.
A common example would be a numerical code that you converted in your analysis to be treated as categorical, but the dataset stored in Construct still sees it as numeric.
For example a saved model could have a memorized formula that says, multiply “word” by 3. This would produces invalid results – expressing a null.
To ensure that data types within the scoring job are correct, open your transform, find out which variables are being treated as numbers and which are text.
Then, add a type convert prior to the transform to ensure that each variable that is supposed to numeric is either integer or real, and each that is supposed to be text is text.