I'd say this is not usable directly, though it can influence how you conduct future research.
See parallel construction for a similar problem in law enforcement. Certain evidence may not be admissible in court, but a law enforcement officer can use knowledge of the inadmissible evidence to find admissible evidence. (Note that I am not saying this is right or wrong of law enforcement to do, just that it happens.)
How this might work in research is that you observe a situation in your personal life, but obviously it would be scientifically and ethically questionable to use this evidence. You can avoid the problem entirely by conducting a newproper study, using your previous observations to help design the study.