Allow people to pre-commit money on a once-off or monthly basis. Then, at the end of a preset period, process that donation in proportion to the amount of time you were meeting or exceeding your goal requirements.

I already donate a regular amount of my salary to charity, and sometimes get a windfall (e.g. tax return) that I donate on a once off basis. However, I would like to be able to stake this money on performance against my goals. If I have a good month, all the money goes to charity. If I derail a couple of times (say, 10% of the month), then 10% (or maybe even more) money is deducted from the full donation. Perhaps you could also earn notionally 'lost' money back by exceeding your goal for a period of time, providing further incentives.

The importance of the precommitment (apart from ensuring that I have the money in the first place) is to tap into our natural sense of loss-aversion - rather than getting the full 'warm fuzzy' of the charitable donation, you get the sting of knowing that you lost money for that charity*. Setting the incentives to run in this direction also means that you don't have conflicting incentives and reduced sting you have with the simple beneficiary model (where you get 'warm fuzzies' when you fail a goal because at least your pledge is going to charity). In fact, because the system is decoupled from pledges, you don't have the issue of needing to change your revenue model (as you would if you were donating your pledge amount). Win-win-win.

While on the face of it might seem that this would decrease the amount of money flowing to charity (i.e. you were planning to donate the money anyway), I think it would incentivise more charitable giving in the first place, because there's double positive reinforcement on offer, meaning that you're likely to continue donating in future.

Where the surplus money goes is up for debate. Obviously the main thing is that it doesn't go to your preferred charity, although anti-charity would be another option (noting the Beeminder team's stated reticence to provide 'anti-charities' as an option).

• I'm sure you're well versed in the cognitive psychology of loss aversion vs gain preference, but for anyone who's not familiar with Kahnemann and Tversky's findings: http://en.wikipedia.org/wiki/Loss_aversion