A scratch on a 3.6-metre toughened shopfront pane is not a small argument. Replacing one on Chiswick High Road means a made-to-measure unit, a glazier with vacuum lifters, a pavement licence from Hounslow, and probably a night shift because you’re not closing that footway at two in the afternoon. Four figures before anybody discusses who caused it. And the answer to who caused it is usually more interesting than either party wants it to be, because in a large share of these cases the scratch was built into the glass months before the cleaner arrived.

What Is Fabricating Debris, and Why Does It Surface on Cleaning Day?

Toughened glass is made by running annealed float through a furnace at around 620°C and then quenching it fast with air, which puts the surface into compression and the core into tension. That’s what makes it strong and what makes it break into cubes rather than daggers.

Inside the furnace the glass rides on ceramic rollers. Those rollers pick up debris – fragments of glass from previous breakages, plus whatever else a production line running all day sheds onto its own rollers – and when a softened sheet passes over a contaminated roller, particles get pressed into the underside surface. They fuse there. The glass cools with tiny hard specks sitting proud of, or partly embedded in, the tin-side face.

You cannot see them. Not from the pavement, not with a torch, usually not at all until something moves across the surface with pressure behind it. Then the particle detaches and travels, and it takes a curved score with it. That arc shape is the signature: a fine crescent, often several of them, running in the direction of the stroke. A blade scratch from a nicked corner is straight and continuous. Debris damage curves.

Fabricating debris is a known and long-recognised issue in glass manufacture, and it is not evenly distributed. Some processors have it under control. Some furnaces are worse in some months than others, and a single shopfront installation can have four clean panes and one that grates under the blade like wet sand.

Ten seconds to work out what you’re looking at

Corner stamps first – toughened glass carries a permanent mark, usually bottom corner, giving the processor and standard. Then step back and look for roller wave, the faint horizontal ripple in the reflection that tempering leaves behind. Polarised sunglasses will show you the quench pattern, a leopard-spot or stripe effect in the reflected light that annealed glass never has. Three seconds each. Write it down before you unfold the ladder.

Does the Scraper Cause the Scratch, or Find It?

Both trade positions on this are defensible, which is why it ends up in dispute.

The physics is settled enough. A tungsten or carbon steel blade is softer than the glass surface it rides on, which is why a clean blade on clean toughened glass leaves nothing behind even after thousands of strokes – the trade has been scraping paint and mortar off shop windows since before float glass existed. Introduce a fused ceramic particle sitting proud of that surface and you have changed the tool. You are no longer drawing steel across glass, you are drawing a small hard point across glass under whatever pressure your forearm is applying.

The North American window cleaning and glass industry bodies reached a broadly settled view years ago: scraping is an accepted method of removing bonded soiling from glass, and where debris is present the resulting damage is a defect of manufacture rather than negligence by the cleaner. UK guidance from the glazing side has historically been more cautious about metal blades on any glass surface, which gives a shopfitter or a glass processor a comfortable position to argue from.

So the technical question and the contractual question separate. Technically, a correctly used blade on debris-free toughened glass does not scratch it. A correctly used blade on contaminated glass will. Contractually, if the specification you signed says no metal scrapers and you used one, none of the physics helps you.

What a defensible scrape looks like

New blade for each shopfront, minimum. Glass wet with solution at all times and never a dry pass. Blade held flat, around five to ten degrees, pushed in one direction rather than dragged back and forth, and lifted clear at the end of each stroke instead of pivoting on a corner. Test patch first, low down in a corner where the sign vinyl covers it, and if you hear or feel that gritty resistance you stop there and photograph the pane.

I stopped a job on Northcote Road once at the third pane and rang the client from the pavement. It cost me the day. It also meant that when the tenant later found scoring on two other elevations, cleaned by somebody else, nobody was looking at me.

Who Ends Up Paying?

Whoever has the worse evidence.

I mean that as a description of how these disputes resolve in practice. Liability nominally turns on cause, but cause is nearly impossible to establish six weeks after the fact when three trades have been on site. What it turns on in reality is whether anyone can show the state of the glass before the cleaning happened, and whether the cleaner can demonstrate they worked to a recognised method.

The strongest position I’ve seen a contractor hold was on a Kingston retail unit where the cleaner had photographed all eleven panes at handover, timestamped, including two with existing scoring near the door furniture. When the debris damage appeared on pane seven, the conversation lasted one email. The glass processor replaced the unit under a fabrication claim. Nobody argued about the cleaning contract at all, because the cleaning contract was never the weak link.

The condition survey nobody does

Twelve photographs and four minutes on the first visit. Every elevation, raking light where you can get it, close on any existing damage, and a note of anything that feels wrong under a test blade. Send it to the client the same day so it sits in their inbox with a date on it.

Almost no window cleaning firm in London does this on retail work, and every single one of them will do it after the first time they eat a £2,800 pane. Write it down.

Why Won’t Your Public Liability Policy Cover It?

Here is the part that catches out competent operators, and it is worth more attention than the rest of the article combined.

Standard public liability cover responds to injury or damage caused to third parties. Nearly every policy carries an exclusion for property in the insured’s care, custody and control – meaning the actual thing you were working on. You drop a squeegee through a display cabinet, you’re covered. You scratch the pane you were sent to clean, the insurer looks at that exclusion and declines.

What you need is an extension variously called treatment risk, damage to goods worked upon, or a care, custody and control write-back, and it is not automatic. It carries its own limit, often far below the main indemnity, and often with a separate excess. Half the cleaning firms operating on commercial contracts in Greater London are carrying £5 million of public liability and £5,000 of the cover that would respond to a glass claim.

Then read the exclusions inside the extension. Several insurers exclude damage arising from the use of abrasives or bladed tools, which is a neat way of removing the only risk you bought the extension for. Others require you to demonstrate compliance with manufacturer cleaning guidance, which pulls the whole GGF-versus-North-American argument back into play at the worst possible moment.

The question to put to your broker in writing

Ask them directly: does this policy respond to a scratch caused by a scraper during routine cleaning of toughened glass, and if so, to what limit. Get the answer by email. A broker who gives you a confident verbal yes and a vague policy schedule is not the person who will be handling the claim.

What Should Go in the Contract Before Anyone Touches the Glass?

Four things, and none of them are unusual asks in a commercial cleaning agreement.

A clause recording that bonded contamination will be removed by blade unless the client specifies otherwise in writing. A clause stating that damage arising from fabrication defects in the glass is not the contractor’s liability. An agreed pre-works condition record forming part of the contract documents. And a notification window – damage reported within seven days or not at all, because after that the site has had other people on it.

That last one gets pushback and it should. Seven days is short. It is also the only version that means anything, because a claim raised in November about a September clean is unprovable in either direction.

The builders’ clean problem, which is a different animal

Most shopfront glass damage in London does not come from routine maintenance cleans at all. It comes from the post-installation clean, where the glass has cement splash, silicone smears, protective film adhesive and mortar runoff on it, and somebody is being paid a fixed price to get it off before handover on a Friday. Alkaline runoff from fresh concrete etches glass permanently within days – that is chemical damage, invisible until the surface dries out, and impossible to polish back. It gets blamed on whoever cleaned last.

If you are fitting out a unit on the King’s Road, the builders’ clean is where your money is at risk, and it belongs in the main works contract with the glazier’s warranty attached to it. Not in a cleaning contract signed afterwards for £340 a month.

I’d tell any facilities manager reading this to price the condition survey into the cleaning tender and pay for it separately, because a cleaner who is doing it unpaid will stop doing it by month four. Sixty pounds a year buys the documentation that decides a four-thousand-pound argument. The glass on that Kingston unit is still in, six years on.