Shearing Layers: How Far to Go on a Retrofit

Shearing Layers: What Duffy and Brand Tell You About How Far to Go on a Retrofit

Assume you've found it: the perfect run-down property — the worst house on the best street — or that problem building in a key location on campus. The planning authority is actively encouraging refurbishment over new-build. Where do you start?

Every retrofit conversation eventually arrives at the same question: how much of the existing building do you actually touch? Full fabric, back-to-brick retrofit, a shell-and-core holding pattern, or the minimum needed to keep the building weathertight while you decide?

This is often assumed to be a budget question first and a design question second. It's usually the other way round.

Where the idea comes from

The concept traces to Frank Duffy's doctoral research at Princeton between 1968 and 1971, on the relationship between organisational structure and office layout, and to the practice he returned to London to co-found in 1971: DEGW, with Peter Eley, Luigi Giffone and John Worthington. Duffy set the idea out fully in writing in Planning Office Space (Duffy, Cave & Worthington, eds., 1976), using an extended theatre analogy across four layers — shell, services, scenery and sets. The shell of a building, he wrote, is the equivalent of the bare stage built to last the life of the theatre; the scenery is the props brought in for a given tenancy; the sets are the day-to-day rearrangement of that scenery as the occupants' needs shift. He returned to the idea at greater length in his 1992 book The Changing Workplace, and later described a building as carrying "a whole series of layers of materials with varying degrees of longevity." Stewart Brand took Duffy's four layers and expanded them into a fuller six-part model in his 1994 book How Buildings Learn: What Happens After They're Built, later adapted into a six-part BBC television series in 1997. Duffy's original terms don't map one-to-one onto Brand's — Duffy's single "Shell" covers what Brand later split into structure and skin, and Brand added "site" as a layer of its own. Brand's six layers, with Duffy's originals noted alongside where they correspond:

• Site — the plot itself; effectively permanent. (No Duffy equivalent — Brand's own addition.)
• Structure — the frame; 30 to 300 years. (Part of Duffy's "Shell.")
• Skin — the external envelope; renewed every 20–50 years. (The other part of Duffy's "Shell.")
• Services — plumbing, electrics, HVAC; replaced every 7–15 years. (Duffy's "Services.")
• Space plan — internal layout; shifted every 3–30 years as use changes. (Duffy's "Scenery.")
• Stuff — furniture, fittings; changes monthly. (Duffy's "Sets.")

The useful insight isn't the list itself — it's that a building in trouble is usually a building where a fast layer has been lost, or embedded in a slow one. Services routed through structural zones. A space plan that can't change without touching the skin, or changing structure. Retrofit decisions that ignore this tend to solve today's problem by making tomorrow's more expensive, complex and difficult to deliver. Whilst this might be sound if that decision is made with eyes open, perhaps set against a limited forward plan or budget, but when the pathway is chosen without a route map, this makes deep retrofit ever more difficult

Where most briefs actually start

In practice, the request we hear most often isn't any of the three tiers below. It's simpler than that: rearrange the rooms, refit the space, make it work for how we use the building now. In Duffy's original terms, this is a request that touches only the scenery and the sets — the space plan and the stuff — and leaves the shell and the services untouched.

There's nothing wrong with that request on its own terms. It's usually the cheapest option, the least disruptive, and the fastest to deliver, and plenty of buildings genuinely just need their layout rethought. The trap is that it's very rarely presented, or understood, as a deliberate decision to defer the structure and skin question. It's simply assumed that touching the fast layers is "the refurbishment," full stop — and the building's actual energy use, fabric performance, and embodied carbon in the slower layers never enter the brief at all.

That matters for two reasons beyond the obvious one of a missed efficiency opportunity. First, a space-plan-only refit can quietly make the eventual proper retrofit harder: new partitions built without regard for future insulation routes, or services re-routed through zones that will need to be opened up again in five years' time. Second, and more simply, it's a decision with a cost attached — a deferred one, in carbon and in energy bills — even though no one signed off on deferring it.

So before the three tiers below become the relevant question, it's worth asking the client directly: is this a fit-out, or is this the retrofit? If the answer is genuinely the former, that's a legitimate choice — as long as it's made with eyes open. If the brief has quietly become "make it feel refurbished" without anyone deciding whether the fabric and structure need attention too, that's the moment to slow down and ask the harder question before the fit-out contractor is on site.

Three tiers, and when each one earns its keep

Once the structure and skin are genuinely on the table — not deferred, but actively part of the brief — we tend to frame refurbishment options as a spectrum rather than a single choice:

Full refurbishment to contemporary performance standards.
Full insulation and airtightness, with structure brought inside the thermal line, new low-carbon services (MVHR, ASHP, PV), and new passive solar shading. This is the right call when the client is holding the asset long-term and the operational cost of a half-measure — energy bills, occupant comfort, maintenance — will outweigh the extra capital spend within a reasonable payback horizon. It's also the only tier that gets you to EnerPHit or equivalent.

Partial refurbishment — skin, shell and core.
Structure insulated and brought inside the thermal line, but services kept minimal or deferred. This suits a client who needs the shell problem solved now — to stop a building depreciating, or to make it lettable — without committing capital to a fit-out for a use that isn't yet defined. It's a bet that the space plan and services will need to flex again soon, so you don't want to lock them in yet.

Minimal repair and secure.
New or deep repair to the weathertight envelope, secured but not upgraded, explicitly designed to be improved later — insulated from below or inside, services added incrementally. This is the right answer when the brief itself isn't settled: an asset held pending a decision on future use, or a phased project where spending on services now would be wasted if the space plan changes in eighteen months.

The carbon question sitting underneath every tier

Money and design aren't the only variables in play — carbon cuts across both, and it pulls in two directions at once. Every measure that improves operational performance — new insulation, an ASHP, better airtightness — saves carbon over the building's future life. But delivering that measure often means removing existing fabric, systems or components, and everything already built carries embodied carbon: the energy and material cost already spent putting it there. Strip out a serviceable but under-performing element to hit a fabric target, and you're paying an upfront carbon cost for a saving that only pays back over years or decades. This is a long-term decision.

This is where the shearing-layers model earns its keep as more than a scheduling tool. The slow layers — site and structure — carry the most embodied carbon and are the most expensive to replace, in carbon terms as much as cost; retaining them is close to always the right call. The faster layers — services, space plan, stuff — carry comparatively little embodied carbon and are where operational savings are won, so they're the layers where replacement is cheap in carbon terms even if it's disruptive in cost or programme terms. The skin sits awkwardly in between: retaining original brickwork or masonry keeps embodied carbon in place, but an under-performing envelope is often where the biggest operational losses are happening, so it's usually the layer that most needs genuine scrutiny rather than a default answer either way.

Read against the three tiers: full refurbishment accepts the highest embodied-carbon cost of intervention in exchange for the largest operational saving, and needs a long enough hold period to justify it. Minimal repair-and-secure keeps embodied carbon almost entirely intact but banks little or no operational saving yet — which is fine if the building's future use, and therefore its real performance target, isn't settled. Skin, shell-and-core sits between the two. None of the three tiers is inherently the lower-carbon choice; it depends on how long the client will hold the asset and how it will be used, which is really the same test that decides the tier on cost and design grounds. Carbon isn't a separate calculation bolted onto the decision — it's the same decision, looked at from a different angle.

The mistake we see most often

Clients — and, to be fair, some architects — often choose a tier based on available budget this year, then discover the building's fast-moving layers (services, space plan) are now locked against its slow-moving ones (structure, skin) in ways that make the next phase of work harder and more expensive than if nothing had been done at all. A minimal repair-and-secure approach done properly should make the eventual full refurbishment easier, not harder. That's the actual test of whether a "temporary" retrofit tier has been specified well: does it keep the slow layers legible, and the fast layers replaceable, or has it just deferred the same problem at higher eventual cost — or made it much harder to rectify later?

So, back to that run-down property or problem building. Where to start isn't the budget, and it isn't the design brief either — not yet. It's reading the layers: which one is actually driving the need to act, how much embodied carbon sits in the layers you'd be disturbing, and whether your chosen tier lets the other five keep moving at their own pace. Get that read right, and the budget and the design brief follow from it. Get it wrong, and you'll be back here again sooner than the client expects — at higher cost, in every sense, than doing it properly the first time.

Images:
1. Illustration from Stuart Brand, How Buildings Learn, pg.13, and widely available on the internet.
2. 'Planning Office Space' (1976) a key source, and as far as we could find, the first mention of the idea of building layers - specifically for office, using the theatre idea, but extends to all types of building, whatever the function.
3. Rodbaston Green Energy Centre, in a retrofit of a granary barn. more details here:

Getting the attribution right took a bit of digging — and a lot of generosity from a few people, so thank you to Ellie Duffy, for pointing us toward the right source when the trail went cold, and to Hiral Patel of the Welsh School of Architecture, Cardiff University, (and formerly of the DEGW archive at the University of Reading) for the context that pinned it down properly.

Sources:
Duffy, F., Cave, C. & Worthington, J. (eds). (1976). Planning Office Space. London: The Architectural Press.
Duffy, F. (1992). The Changing Workplace. London: Phaidon Press.
Brand, S. (1994). How Buildings Learn: What Happens After They're Built. New York: Viking Press.
Brand, S. (writer/presenter). (1997). How Buildings Learn [TV series]. BBC.
Duffy, F. (2008). Work and the City. London: Black Dog Publishing.
Harris, R. (2026). 'Frank Duffy: Researcher and Practitioner'. Buildings & Cities, 15 April 2026. Available at: buildingsandcities.org/insights/news/frank-duffy-researcher-and-practitioner.html