I would track the actual interaction in the CRM system and then select a survey that matches it. Almost nobody does this. So one does not need to ask the customer questions for which answers already exist and turn off the customer ("did you use our website?") or that aren't applicable to the interaction (like in my case).