The seven-input additive model behind the tool, the four verdict bands and what each one actually means for the panel in front of you, the two hard-override conditions that supersede the score, the documentation move that protects the operator from a fabricating-debris claim — and the case for soaking longer instead of scraping harder.
What the Scratch Risk Calculator does, in five points:
The tool's job is not to make the decision for the operator. The decision is the operator's. The tool's job is to put the catastrophic-failure mode on the table before the tool touches the glass, in a form the operator can show the customer when the work is being negotiated.
There is a conversation that happens every time a residential window cleaner finishes a panel and the customer walks out a day later and sees a straight silver line running diagonally across the lower-left quadrant of the glass. The customer photographs the line. The customer sends the photograph to the cleaner. The cleaner reads it on his phone in the truck between jobs and his stomach drops, because the line in the photograph is a fabricating-debris scratch and the line is permanent and there is no version of the next sixty days that does not involve a difficult conversation about who pays for a replacement.
The Scratch Risk Calculator is the tool we built to put that conversation in front of the operator before the work starts. It takes seven inputs — glass type, panel age, cleaning method, visible debris on the panel, chemistry rung, wipe surface, and pre-rinse state — and assembles them into an additive risk score with a verdict band. The output is not a guarantee against scratch events, because no model can be, but it is a calibrated estimate of the risk in front of the operator at the moment of the decision. This piece is the methodology behind the tool: what each input is measuring, why the point values are calibrated where they are, why two specific input combinations get hard-override treatment regardless of the score, and how the math reflects what the trade has learned about fabricating debris over the last forty years.
The scratch-on-tempered-glass argument has been going on in the window-cleaning trade since tempered glass became the residential default in the 1980s. The argument goes roughly like this. The glass manufacturer position is that tempered glass should never be scraped, period, because the tempering process leaves microscopic glass shards fused to the panel surface — fabricating debris — and any hard tool dragged across the panel will pick up a shard and scratch the panel. The IWCA position (the International Window Cleaning Association) is that the cleaner is not at fault for a fabricating-debris scratch because the debris was on the panel before the cleaner arrived — the cleaner is the trigger but not the cause, and the manufacturer or the installer who failed to inspect for debris is the actual defendant. The customer position is that whoever was the last person to touch the panel is the person who pays for the replacement. None of those three positions is going to be resolved at the level of a calculator, but the calculator can put a number on the risk before the work starts so the operator can make an informed choice about which side of the argument they want to land on.
The other thing the calculator does is name the catastrophic failure mode out loud. The argument above is about steel scrapers on tempered glass. That is not the only scratch mode. An abrasive pad on a coated panel scratches the coating before it scratches the underlying glass. A paper towel on a debris-laden panel drags grit through every wipe stroke and produces fine surface scoring that is visible under raking light. A reused cloth from the truck — history unknown, embedded contamination unknown — can do more damage on a single pass than a scraper would on twenty. The tool's job is to surface all of those modes, weigh them honestly, and produce a verdict that reflects the actual risk on the panel in front of the operator.
The risk math is the sum of seven point contributions. Each input is calibrated against the documented incident rate for that input in isolation; the additive model reflects the empirical observation that scratch modes operate substantially independently of each other.
Glass type is the substrate variable, and it carries the second-largest point spread in the tool (range 4 to 20). Annealed float glass — the older single-pane substrate, most pre-2000 residential — sits at 4 points; tempered, the modern safety-glass default, sits at 20. Laminated and heat-strengthened sit in the middle (12 and 14). Unknown glass is treated as 18 — close to tempered but not quite — because the conservative default for an unmarked panel is to assume the high-risk substrate.
The spread is calibrated against the fabricating-debris literature: the catastrophic silver-line scratch is essentially a tempered-glass-only failure mode. Annealed glass does not carry the manufacturing debris that produces the silver line. Laminated panels usually have an annealed outer ply (lower fabricating-debris risk) but are more often exposed to mortar and paint splatter in their typical installation contexts (storefronts, vestibules, security applications) which raises a different scratch mode. Heat-strengthened glass falls between — it goes through a tempering cycle, but a shorter one, with less debris deposition.
Panel age is the temporal variable, and it gates the tempered-glass risk by the weathering effect. A new tempered panel — under three years old — carries its post-manufacture fabricating debris in essentially undisturbed density, and 12 points fall on the panel age input alone. A mid-life panel (3 to 15 years) has had enough rain, wind, hose-rinsing, and routine maintenance to drop most of the loose debris — 5 points. A mature panel (15+ years) has been thoroughly weathered, and the surface debris is essentially gone — 2 points. The reported fabricating-debris incident distribution clusters heavily in the first five years after installation; the panel-age variable encodes that pattern in the additive model.
Cleaning method is the highest-leverage single input in the tool (range 1 to 25), and the calibration reflects how much of the documented scratch incident corpus traces back to method selection. Squeegee-plus-cloth — the standard residential and commercial protocol — sits at 1 point, essentially zero contribution. Cloth-only methods sit at 4 (slight elevation on contaminated panels). Plastic scrapers sit at 6 (real but recoverable scratch potential). Abrasive pads sit at 22. Steel scrapers — razor blades and equivalent — sit at 25, the maximum single-input contribution.
The 25-point allocation on the steel scraper is deliberate. A steel scraper on a clean annealed panel is essentially scratch-free; a steel scraper on a tempered panel with silvery flecks visible is a near-guaranteed scratch event. The point value reflects the worst-case path through the input combinations; the actual risk depends on what the other inputs are doing alongside it, which is what the additive model is for.
Visible debris on the panel is the surface-contamination variable, and it ranges from 0 (clean panel, no contribution) to 24 (silvery flecks visible under raking light). The silvery-flecks reading is the fabricating-debris signature: when you tilt your head and let light rake across the panel at a shallow angle, the loose glass shards from the tempering process catch the light and produce visible specks. The 24-point allocation reflects that this reading is the precursor condition for the catastrophic scratch event; on a panel showing flecks, the next hard tool to contact the surface will pick up a shard and drag it.
The 16-point reading for heavy paint or mortar reflects the post-construction scratch mode — when paint splatter and mortar grit are present on a new tempered panel, the combination produces the second-largest cluster of reported scratch incidents after the silvery-flecks-plus-steel-scraper combination. The 12-point reading for stuck mineral or sap residue reflects the temptation toward harder removal methods that the stuck residue creates: the operator reaches for a scraper to break the residue precisely when the panel is most vulnerable.
Chemistry rung is the lowest-leverage input in the tool (range 0 to 10), because chemistry is not a primary scratch vector — it modulates the conditions that other inputs create. Water alone contributes nothing. Surfactant contributes 1 point. Alcohol-cut mixes contribute 4 (because the faster evaporation shortens the working window and increases the dry-panel exposure). Ammonia contributes 6 (mostly through coating compatibility — softened films come off and expose underlying contamination). Solvent or acidic chemistry contributes 10 (real coating-damage potential on tinted, low-E, or anti-reflective panels).
Wipe surface ranges from 0 (new or laundered microfiber) to 12 (unknown reused cloth from the truck). The spread reflects an underappreciated scratch mode: the wipe surface itself, when contaminated with embedded grit from prior work, becomes the scratch vector. A microfiber cloth that has been used twice and never laundered carries the contamination of both jobs into the next; on a clean tempered panel, the cloth is the most likely source of new surface damage. Newspaper and paper towel sit in the middle (7 and 5) — both are stiffer than microfiber and less effective at grit retention.
Pre-rinse status is the third-highest-leverage input (range 0 to 18), and it is the input most operators undervalue. A flooded panel — water sheeting across the surface — contributes 0 points. A damp panel contributes 3. A dry panel contributes 10. A dry panel with visible debris contributes 18, the largest single-input contribution outside of the cleaning-method variable. The empirical pattern: most reported scratch incidents involve some form of dry-cleaning — spray-and-wipe protocols, dry-cloth wiping, scraper work without pre-flood. The water layer is a lubricant; it suspends grit rather than letting it drag.
Four bands. The thresholds are calibrated against the reported incident distribution.
SAFE fires below 25 points. The combination is in the working envelope where no documented scratch incident has been reported in the published or trade-association incident corpus. The standard checks — raking-light scan, clean wipe surface, flood-rinse on first pass — should still be in place, but the combination does not require special handling.
CAUTION fires between 25 and 49. The combination is in the moderate-risk band where occasional scratch incidents have been reported, mostly under the grit-on-paper-towel mode or the dry-panel-with-light-contamination mode. The adjustment list identifies which inputs are contributing the most points; addressing the highest-leverage one or two usually drops the score back into the safe band without changing the working method.
HIGH RISK fires between 50 and 74. The combination is in the band where the majority of non-tempered scratch incidents originate — post-construction cleanup, dry-cleaning of contaminated panels, paper-towel wiping on debris-laden glass. Do not proceed with the current protocol; the adjustment list orders the changes by leverage, and the highest-leverage one usually drops the score by 15 or more points with a single substitution.
DO NOT PROCEED fires at 75 points or above, or when one of two hard-override conditions fires. This is the catastrophic-failure band; proceeding has a high probability of producing a permanent scratch.
The score-based bands are not the whole story. Two specific input combinations get hard-override treatment because the catastrophe potential in those cases is high enough that the score-based verdict undersells the risk.
Hard-override 1: the fabricating-debris event. Steel scraper plus tempered glass (or unknown glass, treated as tempered) plus visible silvery flecks fires the do-not-proceed verdict regardless of the additive score. The combination is the documented catastrophic-failure mode: the steel edge picks up a fabricating-debris shard, drags it across the panel, and produces a permanent silver line. The replacement cost of a tempered residential panel runs $400 to $1200 installed; the cost of pivoting to a plastic scraper, extending the soak time, or documenting customer consent before proceeding is zero. The override exists because operators sometimes look at a 70-point score, see that it is just under the do-not-proceed threshold, and proceed anyway. With the silvery flecks present, the score-based threshold is not the right test.
Hard-override 2: abrasive pad on dry tempered or heat-strengthened glass. The same logic applies in a different scratch mode. Abrasive pads on dry coated panels produce surface micro-hazing across the contact area; the haze is visible under raking light and cannot be polished out. Flood-rinsing the panel before pad work substantially mitigates the mode (the pad slides on water rather than scrubbing dry glass), so the override only fires when the pre-rinse input reads dry or dry-with-debris.
Neither override fires on annealed glass. The catastrophic modes are tempered-glass and coated-glass specific; annealed substrate is not exempt from scratch risk, but the failure modes are different and the score-based bands handle them adequately.
The high-risk and do-not-proceed bands both produce a documentation recommendation: a before-work photograph of the panel under raking light, sent to email for timestamp, showing the visible debris or silvery flecks before any cleaning work begins. The recommendation exists because the IWCA fabricating-debris position protects the cleaner from liability if the debris was on the panel before the work started — but only if the documentation exists to prove it.
This is the move that most operators do not make and that protects them when the customer photograph arrives a day after the work is complete. Sixty seconds, phone camera, email send for timestamp. The operator who has the before-work photograph wins the conversation; the operator who doesn't loses it.
The tool cannot replace the operator's judgment about the specific panel in front of them. The seven inputs capture the major risk factors; they do not capture every variable. Some examples of what the tool does not see: the texture of the rubber on the squeegee (a worn rubber that is dragging grit is itself a scratch vector); the specific solvent in use (citrus solvent is gentler than mineral spirits, which is gentler than xylene, but the tool aggregates all of these into one chemistry-rung input); the customer's tolerance for scratch risk (some customers explicitly accept the risk on stuck-residue removal; others do not); the operator's experience level (a 20-year operator with a deft scraper hand is at lower risk than a six-month operator with the same tool).
The tool is the floor on the conversation, not the ceiling. A senior operator may legitimately push past a caution-band reading because they know the specific panel; a homeowner may legitimately stop at a safe-band reading and call a professional anyway. The score is the math; the decision is the operator's.
The thing the tool will not let the operator do is proceed with the catastrophic-failure combination without knowing that it is the catastrophic-failure combination. That is the point.
Mara Whitfield is Senior Editor at Window Washing Guide and covers the chemistry-and-damage editorial beat, with particular attention to substrate-specific failure modes. Editorial content is researched and reviewed in collaboration with the Giordano Inc. editorial team, with field input from practicing operators in regions where the failure modes have been documented, plus published trade and materials-science references.