After the product annotation or applying the annotation model to the product data you should visit the Annotation Values view.

Structure

In this view you can survey all annotated and/or predicted values in the two left colored columns.

The first dark-blue column contains all user-annotated values. The second light-purple column contains all predictions by the model. Note that this column is empty if no model was created or applied to the data.

The contained values are colored-coded depending on their state. All colors but black are only displayed in the prediction column as an easy way to compare the annotation and prediction value.

• Black:
  It is a user-annotated value (in the far left column) or a value coming from the data source.
  
  
• Green:
  User-annotated and predicted values are equal → it is a match
  
  
• Red:
  User-annotated and predicted values are unequal → it is a mismatch
  
  
• Purple:
  Product values which are only predicted by the model with no user-annotated counterpart

All these values can be filtered via the four buttons above the table. For example, if you click on "Matches" you only get the product rows where user-annotated and predicted values are equal.

Right from the colored columns, other columns from the data feed are displayed. Per default these are a) Mapped core attributes or training attributes from the model. You can change the column visibility by using the burger button icon ☰ in the top right corner of the table. All visible columns are checked. 

At the bottom you can download your product annotations and predictions as a CSV file.