Didn't read the small print. Actually I just didn't read the regular text right below where I was clicking.
I had a scheduled period of disconnected power at my house this month on 03/08 so on the night of 02/08 I changed the hour of power to 6am to 7am and set up all stuff like heating to warm the house before the power cutoff at 8am. On the night of the third I changed it back to my normal 9am to 10am free hour. Of course I didnt read that it changes for the very same day, not the following day like I thought it would, so what I ended up with was 6am to 7am on 02/08 when I hardly used any power and 9am to 10am when the power was disconnected, resulting in 0% power saved over two days instead of what would have probably been 12%ish. It's my fault for not reading it clearly but I think allowing someone to change an hour of power for a time that has already passed is a bit odd, especially when the graph never shows the current date and the final dates that are on there are always estimates.
I think long enough has passed now to realise that I should stop looking at their website because the data cannot be trusted until the bill is produced, Im either having to point out to them when I had no power or the usage reporting changes from estimated to actual at unknown and inconsistent times. Currently I think I would have been better sticking to Genesis.
Im sure there was someone who said that they self monitor their power consumption, And their system analyses their daily usage and sets the hour of power to whatever allowable time slot had the highest usage.
Also not all smart meters seem to be able to report power disconnection and reconnection times. So a power cut just appears as missing data, as the system isn't smart enough to figure out why the data is missing. Hopefully EK are also taking register reads. (remote readings of what the current total usage reading is). And then adding up all of the 1/2 hour usage data and comparing to the register reads. Meaning that any over billing or under billing should get automaticly flagged.