It turns out that physically monitoring water usage is not that hard. There are a number of water sensors available on the market that are suitable, particularly if they are for experimental purposes, rather than for use in a product. The main issue is legal rather than technical, as at least in NSW, all water carrying equipment needs to be certified as safe for drinking water. Some spare parts on the market have not been tested, and therefore may not meet local requirements.
Jaycar sells a suitable Digital Flowmeter which uses an internal impeller and a magnet. The version sold by Jaycar uses a Hall Effect sensor, although other places sell the unit with a magnetic reed switch sensor. The hall effect sensor is liable to be more reliable, but uses more power. This is only an issue where power consumption is a major issue, such as where things are battery powered.
In my case, I installed two of these sensors, one in line with the incoming cold water into the house, and another on the hot water heater. Since the hot water heater is a mains pressure unit, I installed the sensor on the cold water inlet for the hot water so that the sensor did not heat up. I summised that even though the sensor is rated to 80C, and hot water is normally under this, feeding in water at about 20C would most likely extend the life of the sensor significantly.
The software and sensing architecture is described elsewhere on this web site. The sensor puts out one pulse for every 3-4mL, depending on flow rate and mounting. The output is a digital signal that just needs to be monitored by some type of input sensor. In my case, the input is fed into some RS485 Digital I/O devices mostly used for controling the lighting system