Modelling Assumptions to Question

Every energy model rests on assumptions, and most of them are invisible in a headline result. The model assumes a heating pattern, room temperatures, hot water use, weather, system efficiency and the thermal properties of materials it cannot see inside. Those assumptions are necessary, because no model can know everything about a home, but they should be visible, stated and open to question. A model whose assumptions are hidden is harder to trust than one whose assumptions are listed, even if the hidden-assumption model gives a more flattering number.

Some assumptions are set by the method itself. A standardised rating method fixes occupancy and heating patterns so homes can be compared, and a modeller should not quietly change those standards to improve a result. Other assumptions are the modeller's judgements where evidence is missing, such as the construction of an unseen wall or the efficiency of an ageing boiler. The two kinds deserve different questions: method assumptions should be checked for correct application, while judgement assumptions should be checked against evidence.

Heating assumptions are often the most consequential. A model may assume the whole home is heated to a standard temperature for a standard number of hours, while the real household heats only some rooms, or heats for longer, or keeps the home cooler. None of those real patterns makes the model wrong for its comparative purpose, but each of them changes what the model implies for bills. Hot water assumptions work the same way: standard demand suits comparison, while a household with very high or very low hot water use will diverge from it.

Material assumptions matter where construction is unseen. A wall U-value, which measures how easily heat passes through, may be taken from a table for the home's age and type, or estimated from a partial observation. Either approach is reasonable if stated, but a table value for a well-built example of a type will flatter a poorly built example of the same type. System efficiency assumptions behave similarly: a boiler's rated efficiency is not its in-use efficiency after years of service without maintenance.

Ask for the assumption schedule in writing, and work through it with the modeller. For each major assumption, ask whether it comes from the method, from observation, from documents or from a default, and what would happen to the result if it were different. Pay particular attention to assumptions that favour a proposed measure: if a heating system is assumed to be inefficient, any replacement looks good, so the inefficiency assumption deserves evidence.

Also ask about weather data. Models use a standard weather year or a local climate file, and a home in a windy, exposed location or a sheltered urban site may experience different conditions. Weather assumptions rarely decide a retrofit choice, but they affect absolute figures, and they are a reason not to treat a modelled annual demand as a precise forecast.

Assumption handling differs by method and place. SAP applies its own standard patterns and defaults, and a SAP assessor must follow them rather than substituting household-specific patterns. National methods in EU member states have their own standard patterns, which may assume different room temperatures, heating hours or hot water demand. That is another reason cross-border rating comparisons mislead: the assumptions, not just the buildings, differ.

For a reader, the practical rule is to compare like with like. Two models of the same home that use different assumptions are not two opinions about the same thing; they are answers to different questions. Insist that any revised model states which assumptions changed and why, so the change in result can be attributed to a cause.

A common mistake is to accept a model because its result matches expectations. A flattering result built on generous assumptions is still built on generous assumptions, and it will disappoint when real bills arrive. Another mistake is the opposite: to dismiss a model because its standard assumptions do not match one household, when the model was never meant to predict that household's bills in the first place.

A further mistake is to challenge only the assumptions that hurt a favoured option while leaving favourable assumptions unexamined. Consistent scrutiny, applied to assumptions that help and hinder alike, is what makes a model trustworthy. Selective scrutiny makes it a sales document.

Keep the assumption schedule with the model report, and annotate it with anything you verified or disputed. When a later professional revises the model, that annotated schedule shows them exactly where to start.

Questions to ask include: which assumptions are fixed by the method and which are your judgements, which three assumptions move this result most, and what evidence would change each of them. A modeller who can answer those questions clearly is giving you a usable model.