Models and Actual Bills Reconciliation

Reconciling a model with actual bills means explaining a difference, not erasing it. A model estimates use under standard conditions, while bills record use under real conditions, so a difference is expected, and the task is to apportion it among plausible causes: weather, occupancy and behaviour, system condition and controls, building inputs, and fuel scope. A reconciliation that attributes the whole difference to one cause, without testing the others, is a guess, not an investigation, and it will misdirect any remedy.

Fuel scope is the first separation. A model may cover heating and hot water only, while a bill covers all electricity, including cooking, appliances and lighting, or a gas bill may include cooking as well as heating. Separating fuels and end uses, where meter data allows, prevents the common error of comparing a heating model with a whole-home bill and concluding that the model is wildly wrong. Where separation is impossible from bills alone, that limit should be stated, and the reconciliation should be treated as approximate.

Weather is the second separation. A standard weather year differs from any actual year, and a cold winter will raise heating bills above a standard-year model without implying any building fault. Weather normalisation, adjusting actual use for how cold the period was relative to the standard year, using degree days or a similar method, gives a fairer comparison. A reconciliation that ignores weather will overstate the gap in cold years and understate it in mild years, and it will mislead in both directions.

Behaviour and occupancy form the third separation. A household that heats more rooms, keeps higher temperatures, works from home, or has more occupants than the standard pattern will use more energy, while a household that heats sparingly or is often away will use less. Those differences are normal, not faults, and they explain much of the gap for many homes. Recording heating hours, thermostat settings, room use and occupancy changes during the bill period turns behaviour from a vague excuse into a testable explanation.

System condition and controls are the fourth separation. A boiler that needs servicing, a control that is misprogrammed, a radiator system that is unbalanced, or a heat pump running at high flow temperatures can all raise use above the modelled level for a correctly described building. Checking service history, control settings and system operation is therefore a standard reconciliation step, before concluding that fabric inputs or the model itself are wrong.

Building inputs are the final separation, and they should be checked last, not first, because they are the most expensive to remedy. Wall construction, insulation presence, glazing and airtightness should be verified where a gap remains after weather, behaviour and system explanations are accounted for. Where a building input proves wrong, correcting the model is the remedy, and the corrected model then becomes the baseline for future comparisons.

By place, reconciliation practice is similar across systems, because it is an investigation method rather than a national rule. UK SAP-based models are explicitly standardised, so a gap against bills is expected and is not by itself evidence of a faulty EPC, while EU national-method models are similarly standardised. In both contexts, reconciliation separates standardised from real conditions, rather than judging one by the other without adjustment.

For homes with smart meter data, reconciliation can be finer: daily patterns can show overnight baseload, heating start times and weekend differences that annual totals hide. For homes without such data, annual reconciliation with weather and behaviour notes remains useful, provided its coarser resolution is acknowledged.

A common mistake is to reconcile a single month, especially a cold or mild month, and to generalise from it to annual performance. Another mistake is to reconcile bills that include estimated readings as if they were measured, when an estimated period can distort the comparison for months until a true reading corrects it.

A further mistake is to use reconciliation to justify a favoured remedy, such as attributing a gap to fabric in order to sell insulation, or to behaviour in order to avoid building work. A methodical reconciliation tests each explanation in turn, with evidence, and lets the evidence point to the remedy.

Keep the bills, the weather notes, the behaviour notes, the system check records and the model inputs together as the reconciliation file, with dates. That file lets a later professional test the reconciliation rather than repeating it from scratch.

Questions to ask include: for which fuels and period are we comparing, what was weather and behaviour during that period, and what remains unexplained after those are accounted for. Those questions structure any reconciliation well.