But you can already do it yourself customer side - keep track of what you spend per category, either in a text file, spreadsheet or dedicated budgeting program (Quicken, MS Money, etc).
I've done exactly this for the last few years. I have a bunch of text files with likes like
# 3rd (of march)
mail 2.23 mailing for bookmooch
lunch 6.04 soup and sandwich @ riverfront cafe
espresso 1.29
# 4th
groc 29.57 groc run
etc. I was previously totaling it with a short Python script, but now I'm using a tiny awk script (it's exactly the sort of thing it was designed for). Either way, it's incredibly easy to parse simple text files like that, dump the results out to gnuplot, etc. Just plugging in receipts once every week or two doesn't take very long.
Of course, but it is a PITA and a small minority does it. Still, my point is the complexity problem he addresses in that example is due to the monolithic nature of the solutions he envisioned.
It would seem to me that forcing said person(s) to relocate geographically to carry out that work would cut down a large percentage of the actual exploits. Not an all or nothing result, but largely beneficial, nonetheless.
