Hello! I am about to close out a Qualtrics survey that has 1500+ respondents. Because the survey was intentionally designed to allow respondents to save their progress and return to the survey after time away, some respondents (not surprisingly) never fully completed the survey. I'm wondering about survey "best practices" and whether I should determine a threshold of completion I will accept and then delete responses that fall beneath that threshold, or if there is good reason to keep more incomplete responses?

The last section of the survey is all optional demographic questions, that while useful for analysis are not at all necessary. However, right before the demographic questions are some important questions about the overall impact of the phenomenon I'm studying, so I'm tempted to set a 90% complete threshold and ignore or delete any responses that don't answer the "big picture" questions (but include cases that didn't answer the demo Qs). For methodological legitimacy, do I just need to be consistent and justify my decision making explicitly? Anything else I should consider? (If relevant, the survey was quite long so fatigue likely played a role in dropout, too.)

Thanks for any advice you can pass my way!