The recipe trap: why you can’t retrofit one building and copy-paste the rest
“We did detailed surveys on five buildings and we’ll extrapolate the findings to the rest…”
“…Our retrofit playbook is simple: rip out the gas boiler, install a heat pump, job done.”
Do either of these sound familiar?
If you work in climate consulting or manage a large property portfolio, you’ve probably heard (or said) something similar. We get it, deep building assessments are expensive and time consuming. It’s tempting to resort to the quintessentially human mental model of looking for patterns, archetypes, or shortcuts that can scale up retrofit decisions.
But this approach simply doesn’t work, and certainly not for commercial buildings.
Buildings aren’t one-size-fits-all, but they also aren’t one-of-a-kind
You’d be surprised to learn how often we hear of one-size-fits-all playbooks being deployed across a portfolio. But it’s important to note that buildings aren’t one-of-a-kind either. Our analyses indicate that buildings fall in the messy middle of these two mindsets. You can absolutely detect meaningful patterns and groupings, especially if you’re working with rich data, but commercial buildings in fact vary a lot.
Here are some common examples:
Same city, different roof types.
Same EPC rating, different actual energy use.
Same building archetype, totally different heating fuel, glazing, materials, or layout.
Such variability across portfolios is near-certain, and it has real consequences: you can’t apply a single retrofit strategy and expect consistent or optimal results.
Figure 1: A recent Building Atlas analysis identified high variability in energy use intensity (kWh/sqm/year) in a sample of UK schools
You Don’t Need Archetypes – You Need Attributes
Rather than classifying buildings into a handful of archetypes, we treat every building as a collection of hundreds of attributes: size, shape, location, materials, roof pitch, orientation, boiler type, fuel use, lighting systems, glazing, internal layout, occupancy pattern, and more.
That might sound like overkill, but this detail matters. Many of these features don’t correlate in obvious ways – a large building isn’t necessarily more energy-intensive; a modern façade doesn’t guarantee good insulation; and the way that each attribute interacts with the other also varies.
So when someone says “this one worked well, so let’s do the same on the rest,” they’re flying blind.
The Retrofit Menu vs. the Building Recipe
Where does this leave us? We propose a better metaphor: effective retrofits work like a full-course restaurant.
First, you’ve got a fairly stable menu: a set of known interventions like insulation, ventilation upgrades, LED retrofits, solar PV, or switching to heat pumps.
But every building has different dietary requirements – the specific combination of building attributes that define how it performs.
So while the menu is comparatively static, the best combination of courses changes for each diner (or building). The same measure (say, a heat pump) might taste great (deliver huge savings) for one building and almost none in another – even if they look similar on paper.
…And to figure out which measures actually work where, you need a recipe that adapts. That means modelling, not guessing.
The Role of AI and ML: Modelling at Portfolio Scale
To do this at scale, you need a robust, multivariate model that can process dozens or hundreds of building attributes against their unique combinations. This allows us to predict things like:
Energy use intensity (EUI)
Total energy demand
Feasibility of different retrofit measures
Energy savings, carbon impact, and payback periods
EPC rating improvements
This is where AI and machine learning shine – not to replace human judgment, but to scale it across entire portfolios and direct the manual human labour to the most productive ends.
Instead of surveying 10 buildings and hoping they represent the rest, we simulate performance for every single asset using real data. The result? Actionable insights that reflect the messy, nuanced reality of buildings.
Don’t Fall for the Extrapolation Shortcut
The one-size-fits-all retrofit strategy is a seductive idea – but in practice, it leads to:
Underperformance, when a measure just doesn’t work for a particular building. Good luck installing a heat pump in a building without the electrical capacity – or replacing a façade in a conservation area.
Over-engineering, when a heavy-handed solution is applied where a light touch would do. For instance, if your goal is to nudge a building from an EPC rating at low-C to a B, you might not need full electrification or a deep fabric retrofit. A smaller, more cost-effective measure might get you there faster – freeing up capital for buildings that actually need deeper investment.
Wasted time, because even if you have a “universal” recipe, you still need to assess every building to see whether it applies, and how. That’s not a shortcut. It’s just a slow, manual version of what a smart model could do for you instantly.
Worst of all, extrapolating slows down decision-making across your whole portfolio. Time is money. And when it comes to Net Zero commitments, time is running out.
We developed Building Atlas to make this process less effort upfront. Simply put, you give us the addresses, and we do the rest – running deep, multivariate analysis on each building to identify the right measures, at the right time, for the right outcome. From there, human efforts can be directed most impactfully.
Curious how different your buildings really are?
We’ll show you – just get in touch. Or better yet, throw your “retrofit recipe” at our model and see what sticks.
Olga Khroustaleva is co-founder at Building Atlas
