TL;DR: Keep it simple, fail early, fail often, take notes, improve, repeat

If I had to it all over again, I would do some things differently.

• I would define a minimal viable product (MVP). I would keep any non-essential feature on a separate list and only implement it only if it becomes apparent that it is absolutely necessary.
• I would also start and a failure mode and effects analysis (FMEA) early on. I did it after the fact, and it made some weaknesses obvious. Having that insight earlier would have helped to improve the product.
• I would push to have prototypes as early as possible. I would deploy the prototypes in the lab, office space, and at home to get early experience in deploying the sensors. It would be valuable to include these trial deployment in the FMEA. 
• Initially, I was quite afraid of failures. Now, I don’t mind failing as long as it is early on in the prototyping phase. Failures and their correction make the whole network more robust and help to build confidence. Also, addressing issues in a controlled environment is much easier than contacting case study participants and schedule a visit. Hence, I would try to fail as early as possible and try to predict failures and their probabilities.