I like Dave's response regarding diet and exercise. But, I'd like to add my observations regarding the use of finger-prick glucose monitoring to put that in the record, perhaps for other conversations.
I found that if I apply two test strips (they are not cheap) to one drop of blood, there is no repeatability: one says 126 and the next (same blood) says 145. I've been told that to measure glucose response with any degree of confidence, you need to take several (undefined) readings and average them.
My thought is that, at, say, an average of $1US per reading, and taking, say, half dozen readings per session to gather before and after response data, not to forget finger pain, that's getting expensive in terms of carrying out many repeated experiments.
That drives me in the direction of continuous, or, I think in the case of some monitors, "flash" monitoring.
I do not see accuracy in terms of the relation between monitor readings and those taken by a lab as nearly an important issue as that of the shape of the curve which reflects glucose response.
Those are my thoughts and experiences thus far. I'm looking forward to deeper insights through these conversations.