Building fabric

What is thermal bridging, and how do y-values and psi values work?

A U-value tells you how much heat passes through a wall, roof or floor. But walls don't just have middles. They have corners, lintels, window reveals and places where they meet floors and roofs, and that's often where heat escapes more easily.

These weak spots are called thermal bridges. They make a difference to a home's heat loss, and a bigger difference the better insulated the house is, because the walls themselves are losing less.

For a heat loss calculation, bridging is handled in one of two ways: junction by junction with psi values, or with a blanket allowance that most people call a y-value. Here's what those figures mean, where they come from, how CIBSE 2026 and the MCS calculator each apply them, and why it matters when you're sizing radiators and heat pumps.

The short answer

A thermal bridge is a part of a building where heat gets out more easily than through the construction around it: a concrete lintel over a window, a gap in the insulation at the eaves, the junction between an outside wall and the floor.

Two ways of accounting for the extra loss:

• Psi value (ψ): the extra heat loss along one junction, in watts per metre of junction per degree of temperature difference (W/mK). Multiply it by the junction's length.
• Blanket allowance (y-value, or ΔU): one figure in W/m²K, the same units as a U-value, that stands in for all the junctions at once.

Detailed junction calculations are rare for existing homes. Most domestic heat loss surveys use a standard allowance set by the building's age and the method you're following. Both CIBSE 2026 and the MCS heat load calculator add that allowance to each element's U-value, so a wall at 1.50 W/m²K is priced at, say, 1.65. The wall hasn't changed; the junctions around it are being charged to it.

Where thermal bridges happen in a real house

Walk round almost any house and you'll find the same suspects.

• Lintels. A steel or concrete lintel above a window or door gives heat a path through the wall, especially where it interrupts the insulation.
• Window and door reveals. Where the frame meets the wall, the insulation may stop short or the cavity may be closed with something more conductive.
• Wall-to-floor junctions. Where a ground floor meets an external wall, the insulation might not be continuous, and slabs and other structure can carry heat out.
• Wall-to-roof junctions. At the eaves, loft insulation sometimes stops before it meets the wall insulation, leaving a cold strip along the top of the room.
• Intermediate floors. Joists, concrete floors or structural supports built into external walls interrupt the insulation.
• Corners. Even with continuous insulation, an outside corner has more outside surface than inside surface. This is a geometric thermal bridge.
• Party wall junctions. Where a party wall meets the external wall or roof, heat can travel along the junction depending on how it's built.

There's another kind that's easy to miss: repeating bridging inside the construction itself. Timber studs through insulation, wall ties across a cavity, joists through loft insulation. These belong in the element's U-value, not in a junction allowance. That distinction matters, because you don't want to count the same heat twice. We'll come back to it.

Bridges, condensation and mould

Bridging isn't only about energy. Where heat escapes more easily, the inside surface is colder. If warm, moist indoor air meets it, condensation forms, and given time, mould. A straight line of mould along a ceiling edge, damp round a window reveal or a cold strip at the bottom of an outside wall are all signs of a bridge. They aren't proof, though. Poor ventilation, too much indoor moisture and other defects produce the same symptoms.

Psi values and y-values explained

Psi values: one junction at a time

A psi value, written ψ, is the extra heat loss along a junction over and above what the U-values of the elements either side already account for. It's in W/mK: watts per metre of junction per degree.

Take a lintel over a window, 1.5 m long, with a psi value of 0.30 W/mK:

At 21 °C inside and −3 °C outside, a 24 °C difference:

About 11 watts for one junction. Not much, but a house has a lot of junctions: every window surround, the floor perimeter, the roof edge, every corner. Psi values come from thermal modelling of the junction's geometry and materials, and they earn their keep in new construction, where the details can be designed and checked before anything is built.

The blanket allowance: all the junctions in one number

Calculating every junction takes time and needs information you don't usually have on an existing house. So instead you use one allowance in W/m²K.

If you did know every junction, the equivalent allowance is the sum of each psi value times its length, divided by the exposed area it's spread over:

That is how an EPC uses it, in SAP: the whole-building heat loss is the sum of U × A for every element, plus y times the whole exposed envelope.

Heat loss calculations for heating design don't do it that way. Both CIBSE 2026 and the MCS calculator add the allowance to each element's U-value and let it ride through the element's own calculation, so a wall facing outside carries its allowance at the full temperature difference, a wall onto an unheated garage carries it at a smaller one, and a tall room's bridging scales with the room. A wall at 1.50 W/m²K with an allowance of 0.15 is priced at an effective 1.65 W/m²K. The wall material hasn't got worse; it's a bookkeeping way of including the junctions.

Point thermal bridges

There's a third kind, the point bridge, written χ (chi) and measured in W/K: localised losses at individual components such as structural fixings or balcony brackets. You won't deal with them separately on a domestic survey, but they exist in detailed building calculations.

Default allowances and where they come from

For existing homes the allowance comes from a table, by age. The one most people have met is RdSAP's, used for EPCs. RdSAP 10, Table 21, gives these defaults for conventional dwellings:

The band is the main dwelling's, not each element's. A 1930s house with a 2015 extension is band C throughout under the standard RdSAP approach. The pattern makes sense: older houses tend to have less effective insulation at their junctions, and as regulations and details improved, continuity got better. But these are defaults, not measurements. Two houses built the same year can bridge very differently, especially if one has been refurbished.

How the MCS heat load calculator applies it

The calculator uses two different sets of figures, and this is where two apparently similar calculations part company.

External elements take the RdSAP Table 21 figure above, 0.15, 0.11 or 0.08 by age band, added to the base U-value to make an effective U-value. That is what the MCS documentation says, and it is what the calculator prints on walls, roofs and floors whose U-value came from RdSAP.

A second, lower set applies to internal elements and to elements facing unheated spaces: 0.10 W/m²K for age bands A to K (built up to 2011) and 0.05 for L onwards (2012 and later). MCS derives these from the CIBSE Domestic Heating Design Guide and publishes them in its own documentation. The documentation also says that a surface with a known U-value typed in, and no bridging factor of your own, takes this lower set rather than the RdSAP figure.

One thing the documentation doesn't say, which we found by reconciling Heatworx against the calculator's own exported workings on three real properties: windows and doors also take the lower set, even though they face outside. The documented rule reads as if every external element takes the RdSAP figure; the calculator's working prints 0.10 on the openings. If you compare against it, expect that.

So for a 1930s house:

• An external wall with a default, RdSAP-sourced U-value gets 0.15.
• The same wall with a known U-value typed in, and no bridging factor specified, gets 0.10.
• The window in that wall gets 0.10 either way.

Across a 30 m² wall at a 24 °C difference, 0.05 × 30 × 24 = 36 W of difference from the allowance alone. Small, but it's the same wall, and it is the kind of thing that makes two surveys of one house disagree.

How CIBSE 2026 applies it

CIBSE's guide takes a simpler line. One allowance for the whole house, chosen by the insulation standard the house was built to, with the main dividing line at 2006, added to every U-value used in the fabric calculation: walls, windows, doors, floors, ceilings and walls to unheated spaces alike, each at that element's own temperature difference. There's no RdSAP age-band table, no external-versus-internal split and no change of default when you type in a known U-value.

Strictly, the guide classifies by the standard the insulation was built to, not the calendar year. In practice a surveyor reads that off the build year unless the house has been thoroughly refurbished, and that is what Heatworx does by default, with an override for the refurbished case. Notice too that MCS's lower set of figures is derived from this table, which is why the numbers are the same, but MCS draws the line at 2011/2012 (the RdSAP band boundary) where CIBSE draws it at 2006. A house built in 2008 gets 0.05 under CIBSE and 0.10 under MCS's internal mapping.

The principle under both is identical: account for the junctions without counting anything already inside the element's U-value. When you compare two calculations, check which allowance each used and how it was applied, not just the total.

Worked example: a 1930s living room

A living room in a 1930s semi: 4 m deep, 3.5 m wide, 2.4 m ceiling. A front external wall with a window, an external gable, a party wall opposite and a suspended timber floor. Design temperatures 21 °C inside, −3 °C outside, a 24 °C difference. Representative U-values for an older, poorly insulated house. This is a simplified illustration: the floor takes the full 24 °C difference, and the same 21 °C is used for both houses below so the comparison is like for like. A real calculation uses the method's own floor treatment and room temperatures.

Before bridging

Fabric heat loss before bridging: 39.3 × 24 = 943 W.

With the allowance

Under the MCS treatment the fabric heat loss becomes 943 + 112 = 1,055 W, with bridging about 11% of it. Under CIBSE it's 1,020 W and about 8%. In a poorly insulated house the walls and windows lose so much that bridging is a modest share either way.

Why it matters more in well-insulated homes

Same room, in a house built in 2015, with walls at 0.18 W/m²K, windows at 1.20 and a floor at 0.13. Fabric before bridging is about 8.1 W/K, so around 195 W against the older house's 943 W.

Now the allowance. Under MCS the house is band L, so 0.08 on the walls and floor and 0.05 on the window: 0.08 × 29 + 0.05 × 3 = 2.47 W/K, which is 59 W at 24 °C. Under CIBSE the house is post-2006, so 0.05 on everything: 0.05 × 32 × 24 = 38 W.

The newer house has a smaller allowance and a far bigger share. That's why bridging becomes more important as buildings improve. It doesn't mean modern houses have worse junctions; a well-designed junction loses very little. It means that once you're using a generic allowance rather than real junction figures, the allowance can be nearly a quarter of what's left. The same thing happens when you retrofit an older house: insulate the walls, improve the windows, top up the loft, and the junctions you didn't touch become a bigger slice of the remainder.

How Heatworx handles it

One allowance, applied the way each method says. Heatworx doesn't take psi values; a domestic survey almost never has them. Under CIBSE 2026 it uses the guide's figure for the property's build year, with the 2006 line, added to every element at that element's own temperature difference, and scaled up with the rest of the fabric in tall rooms. Under MIS 3005-D it uses the RdSAP Table 21 figure on external surfaces and the CIBSE-derived figure on internal ones, by the survey's build year. The report prints which was used.

One override, never silent. You can replace the default with a blanket figure from the guide's rows, for instance where you know a house has been refurbished with continuous insulation or, the other way, where solid ceilings break it. The override replaces the default and never stacks on top of it. Under MCS the app tells you that doing so overrides the MIS age-band treatment and may need justifying for compliance, which is the same warning the official calculator gives for its own overrides.

One known difference from the official calculator. Heatworx applies the external figure to every external surface, windows and doors included. The calculator, as described above, applies its lower figure to openings and to any known-U surface. On the properties we've compared so far the difference has been around 1% of fabric heat loss, with Heatworx producing the slightly higher figure; how much it is on your house depends on how much glazing it has and the U-values involved.

Custom build-ups don't double count. If you build your own wall type in Heatworx, repeating bridging inside the wall is handled in the U-value: a stud zone is declared as a mixed layer with its timber fraction, and mortar joints in a masonry layer are included automatically. The framework's junction allowance is then added on top exactly as it is for a catalogue wall. So keep lintels, reveals and floor edges out of your layers; they're the allowance's job.

Common mistakes with thermal bridging

1. Leaving it out. A smaller share in a poorly insulated house, a large one in a well-insulated one. Omit it and you risk undersizing the heating system.
2. Counting the same heat twice. Studs, wall ties and joists inside an element belong in its U-value. Don't add a junction allowance for them as well.
3. Using the extension's age instead of the main house. The RdSAP approach, and the MCS defaults built on it, use the main dwelling's band.
4. Not noticing that a known U-value changes the MCS default. Type in a wall's U-value and leave the bridging factor blank and the calculator drops from the RdSAP figure to the lower one. Check, especially when two calculations seem to use the same wall.
5. Assuming CIBSE and MCS apply the same allowance. Same numbers for internal elements, different year line, different rule for external ones. Compare the allowance, not the total.
6. Applying psi values to the wrong dimension. A psi value is per metre of junction, not per square metre of wall. A lintel needs its length; a floor edge needs the perimeter.

Quick questions

The bottom line

Thermal bridging is the extra heat that gets out through the junctions and weak spots. Psi values price those junction by junction; a blanket allowance stands in for them when you don't have the detail, which on an existing house is nearly always. Both CIBSE 2026 and the MCS calculator add that allowance to each element's U-value, but they pick it differently and apply it to different surfaces. Know which allowance your calculation used, where it came from, and that the same heat isn't already inside a U-value, and you've got a more reliable picture of what the house actually needs.

Sources and further reading

• MCS Heat Load Calculator, Thermal Bridging
• RdSAP 10 Specification, Table 21
• CIBSE Domestic Heating Design Guide 2026, section 2.5.3.5 (CIBSE knowledge portal, paywalled)
• BRE, SAP 10 Technical Paper S10TP-05: Treatment of Thermal Bridges
• Zero Carbon Hub, Thermal Bridging Guide
• BS EN ISO 6946 and BR 443, for repeating bridging inside an element

Related guides