The major difference between this and traditional degree-day analysis, is that we are using weather data to the hour over the entire history of the meter.
Contrast an overclouded day where night time temperatures are buffered by the clouds with clear skies when days are warmer and nights far colder. Under these "clear-sky" conditions a typical office building will perform better, because little heating is needed at night despite the cold - simply because no ventilation occurs.
Other items of interest are that, although the onset of heating occurs at different temperatures at different times of day (and week), the balance temperature is defined as when "overall" heating is needed at that temperature.
Degree-day analysis is too coarse to capture this reality because with degree-days the whole consumption is grouped against a single load condition. In the diagram above we see a very crisp onset of heating at between 12 and 13 Celsius - and this is regardless of time of day. The balance temperature (necessary for degree-day analysis anyway) can be found with a high degree of accuracy.
Traditional methods for finding the balance temperature systematically assume linear energy use verses load. In our example above it is very clear that at around 1 Celsius load plateaus. This would be seen as a high-load horizontal cluster in a degree-day analysis as shown below:
These two operational plateaus are very vague in this Degree-Day Analysis directly above but are crystal clear in the Outside Air Temperature summary chart at the head of the page that relies on the same input data.
Our claim is not that Degree-Day Analysis is irrelevant, but that given far better processing capabilities and energy metering and weather resolution, sophisticated new pattern recognition techniques (like those that identify balance and limits on the first chart above), are an idea whose time has come.
Appropriate analysis in the situation shown above could save extremely significant sums for the anonymous Supermarket European shown above.