Insights BTW login
← Insights
U-value & condensation 20 Oct 2026 5 min read

A good U-value can still grow mould at the sill. Here's the gap.

A whole-window U-value is an area-weighted average across about seven sections, flattered by glass area, and measured on a standard size nobody actually builds. Condensation doesn't read averages — it finds the single coldest section. Here's the gap between the rating and the real door.

DM By Dave Morehouse, founder, AFRC Lab

I was looking at a sliding door at an industry night recently, and it sent me down a rabbit hole I keep coming back to: how a door can carry a genuinely good U-value and still fog up and grow mould at the sill. The two sound like a contradiction. They're not — and the reason is buried in how the number is made.

A sliding door isn't one number — it's about seven

Thermally, a sliding door isn't one thing. To model it properly to ISO 15099 you run about seven separate sections — the head, the jambs and the sill, each on the inside and the outside, plus the interlock where the panels meet. Each is solved on its own, then they're area-weighted — together with the glass — into a single whole-window U-value. So the headline number you get handed is an average.

And averages flatter. Two things in particular. First, a section that performs badly — a thermal break that's been bridged by a solid sill tread, say — is only a small slice of the total area, so the good sections carry it. Second, a slider is mostly glass. If the glazing's good, it dominates the weighting and pulls the whole-window number down. The U-value can read excellent largely because of the glass, while a frame section quietly runs cold underneath it.

And it's a door nobody builds

There's one more layer. And to be fair to the rating, it isn't cutting corners here — a WERS rating runs the very same sections to the very same standard. It just reports at a standard size — for a sliding door, a standard two-panel door roughly 2m × 2m — so one product can be compared against another like for like. Fair enough. But when's the last time a real project installed one of those? Real doors are bigger, more glazed, stacked or multi-panel. Different size, different glass-to-frame ratio, different area-weighting — a different number. So the figure on the certificate is an average, of a window that doesn't exist on your job.

Rated to compare, modelled to build. Same sections, same standard — one number reports a catalogue window, the other reports yours.

Condensation doesn't read the average

Here's why it matters. Mould doesn't form on a weighted mean. Condensation forms at the single coldest surface on the assembly on a cold night. So a door can meet its U-value target — on a standard size, on the strength of its glass — and still have one section, the sill track or a bridged break, cold enough to sweat and grow mould. The headline never saw it, because the headline was an average of a different door.

That's the whole difference between a number and a model — not the method, the window. A rating reports the generic, catalogue window, so products compare like for like. A model reports the actual door — your size, your configuration, every section — so you can see where it runs cold. One is a comparison label; the other is the window you're about to install. That's the reason for a performance model: it rates your window, not a standard one.

What to actually do about it

Model the real door. Not the standard test size — the size and configuration going into your project, every section to ISO 15099, so the whole-window number reflects the door you're buying.
Screen the cold spots — fRsi. A fast surface-temperature factor on each junction and the sill that flags which section is heading for condensation, so the weak point is visible, not hidden in the average.
Prove it where it counts — TIRS. Where the risk is real, a TIRS assessment establishes the Mould Index — the certifier-ready condensation evidence the NCC asks for — on the section that governs.

None of this makes U-value useless — it's a good measure of the thing it measures. It just isn't the condensation answer, and it isn't your door. Model the real one, section by section, and the cold spot turns up on the drawing board, where it's a glass or a detail change — not at handover, where it's a callback and a mould claim.

Not sure your door is as good as its number? Send us one.

Pick the sliding door — or any window — you're least sure about. We'll model the real sections to ISO 15099, at the size and configuration that actually gets installed, and run the surface-temperature (fRsi) screen so you can see where it runs cold. When you need the compliance evidence, TIRS MI-6 delivers the Mould Index the NCC asks for — while it's still a redline, not a rectification.

Talk to the lab →

About these figures. A whole-window U-value (Uw) is an area-weighted average of the glazing and the frame sections, modelled to ISO 15099 / NFRC 100; it is a comparison measure, not a condensation assessment. Surface-temperature-factor (fRsi) results are an ISO 13788 steady-state screening indicator — they flag risk, they do not certify compliance. Under NCC 2025, Parts F8 and H4 determine condensation compliance from a Mould Index, which a TIRS assessment establishes. Standard rating test sizes are set by the rating scheme and differ from real installed sizes and configurations. Modelled results are indicative to the systems, sizes and junctions provided; final NCC compliance is confirmed by your assessor or certifier. AFRC Lab (Advanced Fenestration Research & Certification) is an independent consultancy; it is not the AFRC and is not affiliated with WERS or the Australian Fenestration Rating Council. BTW™ and TIRS™ are trademarks of AFRC Lab Pty Ltd.