Unfortunately this individual only has 2 years of data, which makes it impossible to get a proper deviation from a base line year. But there is enough data to see the spike in consumption after the smart meter went in.
One of the problems with trying to get a deviation using a previous year is that H1 only took readings every 3 months, estimating the previous 2. This means that the month of the reading has consumption has a proportion from the previous two, or the previous two have a proportion of the next reading. Thus, using an estimate month as a base line is not going to be right unless some rational mathematical method can be used to move some of the consumption off the actual month read. In the case of this individual I experimented with a few possibilities.
The premise is that a percent of the read month should be distributed to the previous estimated months. Judgment is needed to see if that percent makes sense. 15% seemed reasonable, but is still a guess. It may be between 10 and 20% of the read month to be distributed to the previous months. That's assuming the previous month's estimates were low. If high, the game is played in reverse. The question is how does one know if the previous month's estimates were too high or too low? Again it has to be a judgement call.
We are really going to see this with another example once I get all the data as the next person has records back to 2003.
So here is the spreadsheet of the values with the actual read and estimates and the adjusted.
Notice the dramatic difference between Feb 2010 and Feb 2009. 2.14 times higher (up by 114%)
You can clearly see the dramatic jump after the smart meter went in part way through Aug 2009, notice even that month appeared to be too high.