In this case automated phone calls to home lines and surveys to mobile devices. The sample is then "blended" by including the responses from each type into one aggregation. The sample can still be and should be random

And blended sampling does carry risk that single source sampling does not. Pollsters have to establish how each type of information is weighted when putting together the complete sample. In this case an automated phone call and survey, neither of which are my favorite sources, are a good match. Both sources are removed from talking to a real person, which diminishes inherent and subtle answering ques for each question.