In addition to this, I've started sampling days. Instead of recording transactions for every day, I can do something like every fifth day. This reduces precision, but it's still within a range acceptable for my purposes.
(Though it's important that it's either a systematic sample (every n days) or a random one (throw a dice each day, if it's 5 or higher, record that day). The days you pick have to be uncorrelated with the types of expenses those days, or your results will be biased.)
(Though it's important that it's either a systematic sample (every n days) or a random one (throw a dice each day, if it's 5 or higher, record that day). The days you pick have to be uncorrelated with the types of expenses those days, or your results will be biased.)