— reviewed against the regs as they stand.
Run the same wall through two different online calculators and you will often get two different numbers. Sometimes the gap is trivial. Sometimes it is enough to move a design from passing to failing. Either way it is corrosive: if the answer changes with the website, none of them feel trustworthy.
It is not confined to U-values. Discussing thermal-mass tools, contributors on the Green Building Forum found that different calculators produced “very different results for decrement delay” for the same construction — and were left unsure which model to believe.
The reassuring part is that the calculators are usually not broken. A U-value is the output of a method with several inputs, and two tools can each be arithmetically correct while disagreeing, because they made different assumptions about those inputs. There are five places the difference nearly always hides.
1. The bridging fraction
This is the big one, and on a framed wall it will usually account for most of any gap.
Studs, joists and rafters conduct far better than the insulation between them, so the calculation has to know what proportion of the area they occupy. On a timber frame wall at 600mm centres with 38mm studs, the timber is not simply 38 ÷ 600 — you have to include noggins, top and bottom plates, lintels and the extra studs at openings, which is why the figure used in practice is typically higher than the naive one.
Move that fraction from 0.09 to 0.15 and a 140mm mineral wool wall moves by roughly 0.02 to 0.03 W/m²K. That is the whole margin on a 0.18 target.
How to check: find where each calculator states its fraction. If one of them never asked you, it assumed one.
2. Whether the combined method was used at all
Once anything bridges an insulation layer, resistances cannot just be added up. BS EN ISO 6946 requires two separate calculations:
- an upper limit, treating heat as flowing down parallel paths (through timber, through insulation) that do not interact
- a lower limit, treating each bridged layer as a single equivalent layer
The answer is the average of the two. A tool that adds a single column of resistances is not doing this, and on a bridged element it will read optimistic.
How to check: a calculator following the method can show you both limits. If it only ever shows one total resistance and your wall has studs in it, that is your answer.
3. The ΔU corrections
BR 443 adds small allowances to the finished U-value:
- for air gaps in or around the insulation layer, depending on how well it is installed and whether it is continuous
- for mechanical fixings that penetrate the insulation
They are usually between 0.00 and 0.04 W/m²K. Individually small; collectively the difference between a pass and a fail on a tight target. They always make the number worse, they are easy to leave out, and plenty of free calculators leave them out.
How to check: look for ΔU as its own line. If it is not shown, assume it was not applied.
4. Surface resistances and the direction of heat flow
Every element has an internal and external surface resistance, and they depend on which way the heat is going:
| Element | Rsi | Rse |
|---|---|---|
| Wall (horizontal) | 0.13 | 0.04 |
| Roof (upward) | 0.10 | 0.04 |
| Floor (downward) | 0.17 | 0.04 |
Use wall values on a roof and the answer shifts. There is a second trap here too: if the build-up contains a well-ventilated cavity, ISO 6946 §6.9.4 requires the cavity and every layer outside it to be disregarded entirely, with Rse replaced by Rsi. A calculator that quietly includes the outer leaf in that situation will give a much better — and wrong — answer.
5. The λ values themselves
A generic value for “mineral wool” and a manufacturer’s declared value for a specific product are not the same number. Declared values are revised. Some products have thickness-dependent λ, so the same board is 0.027 at 50mm and 0.025 at 100mm. And a fair few online calculators carry values with no stated origin at all.
How to check: every material should show where its figure came from and when it was taken. If a calculator cannot tell you that, its answer cannot be audited — and an unauditable number is a poor thing to attach to a Building Regulations submission.
So which one is right?
Work through the five in order. In our experience the difference is found by the second or third on almost every occasion, and it is nearly always a bridging fraction or a missing correction.
Then apply the harder test: which calculator will show you its working? A tool that prints the method, every layer with its thickness and λ, both resistance limits, the corrections applied and the source of every material can be checked by you, by a Building Control officer, or by whoever inherits the file in five years. A tool that prints a number cannot.
That is the difference that matters, and it is why U-Monkey shows the full
substitution — R = d ÷ λ with the actual figures in place — and puts a verifiable
report ID on every report, including the free ones.
Sources: discussion of calculator discrepancies and thermal-mass tools on the Green Building Forum; material-matching and vapour-unit difficulties on BuildHub. The calculation requirements described here are from BS EN ISO 6946:2017 and BR 443 (2019).
Common questions
Which U-value calculator is correct?
The one whose assumptions match your construction and whose method matches BS EN ISO 6946. A calculator is not correct or incorrect in the abstract — it is correct for a particular set of inputs. Before comparing two results, check that both used the same bridging fraction, both applied the combined method, both applied the BR 443 corrections for air gaps and fixings, both used the right surface resistances for the direction of heat flow, and both used the same λ values. In practice one of those five differs, and once you find it the gap explains itself.
How much difference does the timber fraction make?
More than any other single input on a framed wall. Moving a stud fraction from 0.09 to 0.15 on a 140mm mineral-wool wall typically moves the U-value by 0.02 to 0.03 W/m²K, which is the difference between passing and failing a 0.18 target. The fraction is not a matter of opinion: it follows from the stud width and the centres, plus the noggins, plates and lintels, and BR 443 sets out how to arrive at it. A calculator that does not ask you for a fraction has assumed one.
What is the combined method and why does it matter?
Where anything bridges an insulation layer — studs, joists, rafters — the resistance cannot simply be added up. BS EN ISO 6946 requires an upper limit (heat flowing in parallel paths) and a lower limit (heat flowing through equivalent layers) to be calculated separately and averaged. Simple tools that add resistances in a single column are not doing this, and on a bridged element their answer is optimistic. Where nothing bridges, the two limits coincide and the distinction disappears.
What are the ΔU corrections and why do many calculators skip them?
BR 443 requires small additions to the U-value for air gaps in the insulation layer and for mechanical fixings that penetrate it. They are typically 0.00 to 0.04 W/m²K. They are easy to omit, they always make the answer look better, and many free calculators do not apply them at all — so a tool that shows ΔU as a separate line is not being pedantic, it is showing you something the other tool left out.
Do surface resistances vary?
Yes, with the direction of heat flow. BR 443 uses Rsi 0.13 and Rse 0.04 for horizontal flow through a wall, 0.10 and 0.04 upward through a roof, and 0.17 and 0.04 downward through a floor. Using wall values on a roof shifts the result. A well-ventilated cavity changes it again: ISO 6946 §6.9.4 disregards the cavity and everything outside it, and substitutes Rsi for Rse.
Can the λ values themselves be the cause?
Often. A generic figure for mineral wool and a manufacturer's declared figure for a specific product are different numbers, and datasheets are revised. A calculator that will not tell you where its λ came from cannot be checked. Look for a stated source and a date against every material.
How do I prove which answer is right to Building Control?
Show the working. A submission that states the method, lists every layer with its thickness and λ, shows the upper and lower resistance limits, states the corrections applied and cites the source for each material can be checked line by line. A bare number cannot. Every U-Monkey report carries the full working plus a report ID that anyone can verify.