What the diagnostic does, why the four-question structure exists, the seventeen failure modes it can identify, and the cases where the tool sends you to a working cleaner instead of trying to solve the problem itself. The framework that built it is the same one I learned in trade school.
What the Streak Diagnostic does, in five points:
Diagnosis is the part of the trade that takes the longest to teach a new cleaner. The Streak Diagnostic is the closest we can get to teaching it through a screen.
In Copenhagen, the trade school I went through in the early 2000s had a teaching method for window-cleaning diagnosis that the senior instructor called de fire spørgsmål — the four questions. Before any cleaning chemistry came out, before any tool touched the glass, you walked the property with the homeowner and asked four questions. Where is the problem? What does it look like? When did it appear? What have you already tried? The four answers were almost always enough to identify what was wrong.
The Streak Diagnostic on this site is a structured version of those four questions, ported into software, with the visual archetypes that take the longest to describe in words rendered as inline illustrations the user can point at. It is the closest I can get to walking onto a property with a homeowner and asking them what they're looking at.
This piece is the methodology behind the tool. It is for the user who has run the diagnostic and wants to understand why it asked what it asked, what the seventeen possible verdicts actually mean, and where the framework comes from. It is also, less directly, an argument for why diagnosis is a real skill — one that takes years to teach a working cleaner — and why the Streak Diagnostic is the most ambitious tool we've built.
Every cleaning failure on residential glass falls into one of five high-level categories. The first question of the diagnostic narrows the user to one of these five.
Surface failures — the haze, streaks, or film is on the outside or inside surface of the glass and can in principle be wiped off, scraped off, or cleaned off with the right chemistry. Most cleaning calls are surface failures. Most surface failures are caused by something correctable.
Between-pane failures — the haze is inside the sealed cavity of a double-pane window. No cleaner on earth can reach it. The unit has failed and the conversation is about replacement.
Discrete spot failures — the problem is not a continuous haze but a pattern of distinct droplet marks, dots, or spots. Almost always hard-water deposits, sap, tar, or insect residue. The chemistry is specific to the kind of spot.
Edge and frame failures — the problem is at the perimeter of the pane, the gasket, or the frame. Often mold, often moisture intrusion, often not actually a glass-cleaning problem at all.
Mechanical failures — the problem is a scratch, a chip, or a gouge. The cause is in the past tense; the question is whether the damage is reversible.
The first question of the Streak Diagnostic asks the user to identify which of these five categories their problem falls into. The visual cues for each are different enough that misclassification at this stage is rare — surface haze does not look like discrete droplets which does not look like a scratch which does not look like edge mold. If the user does misclassify, the second question's options are designed to surface the mismatch and let the user back up.
The second question varies by branch. This is where most of the diagnostic work happens, because most of the seventeen verdicts are distinguished by visual archetype rather than by underlying chemistry, and the second question is the visual-archetype match.
For the surface failure branch, the four options are:
The first three are correctable by the homeowner with the right protocol. The fourth is essentially permanent damage and requires either professional polishing or replacement.
For the discrete spot branch, the four options ask about the color, texture, and location pattern of the spots. Chalky white droplet marks indicate hard-water deposits. Gritty deposits with slight texture indicate either calcium silicate or efflorescence (concrete leachate). Amber sticky droplets indicate sap or tar. Black dots specifically on the rubber gasket indicate gasket mold.
For the edge and frame branch, the four options separate gasket mold from wood-frame moisture intrusion from the dark-edge-band that means a failed perimeter sealant from the post-pressure-washing edge haze that's its own particular failure mode. Each one of these has a different fix and a different protocol; misidentifying which is which produces months of cleaning that fails to resolve the underlying problem.
For the mechanical failure branch, the two main options distinguish many parallel hairline scratches in one direction (almost always razor scraping on tempered glass with embedded fabricating debris — see the glass types reference) from a few isolated scratches in random directions (mechanical contact with grit, gravel, or hard tools).
For the between-pane branch, the diagnostic doesn't ask a second question. The verdict is delivered immediately because the diagnosis is sufficient — if the problem is between the panes, no further detail changes what to do about it.
Some branches require a third disambiguating question to separate two diagnoses that look similar at the second-question stage. The third question is always closed-form — yes/no or a brief multiple choice — and is calibrated to break a tie that the second question alone leaves ambiguous.
The most useful third question, the one that breaks the most ties, is the temporal question. Did this problem appear suddenly, or has it built up gradually over weeks or months? Many surface failures look identical when you arrive but have completely different causes depending on whether they appeared overnight or progressed slowly. A sudden iridescent sheen on a kitchen window is almost always cooking oil aerosols condensing in a single heating event; a gradual iridescent sheen on a living-room window is almost always plasticizer migration from nearby vinyl trim. Same visual archetype, different cause, different fix.
Another useful third question is the recent activity question. Has the property been pressure-washed, painted, or had construction work done recently? This question alone identifies several of the seventeen verdicts — detergent residue from pressure washing, paint overspray haze, post-construction grit haze — that would otherwise be mistaken for ordinary cleaning problems.
The third question is not always necessary. Many branches reach a confident verdict at the second-question stage and skip directly to the diagnosis. The framework is built to ask the minimum number of questions that produce a confident answer, not the maximum number that could conceivably refine the diagnosis further.
The current version of the tool can produce one of seventeen distinct verdicts. They are:
The seventeenth verdict — the I cannot tell from here outcome — is the one I am most insistent on keeping in the framework. A diagnostic tool that always produces a confident answer is a tool that is sometimes wrong and is never honest about being wrong. The Streak Diagnostic is allowed to admit that some failure modes don't show up clearly enough on a screen to be diagnosed remotely. When that happens, it says so, and it routes the user to the contact page or to a working cleaner.
The Streak Diagnostic is a triage tool. It is the first ten minutes of a cleaning consultation, ported into software. It is not a replacement for the next two hours, which involve actually walking the property, touching the glass, smelling for solvent residue, running tests with isopropyl alcohol on the affected area, and looking at the property's water source, ventilation, and surrounding conditions.
For most residential cleaning failures, the first ten minutes is enough. The diagnostic produces a confident verdict and a linked article, and the homeowner can fix the problem from the article. For some failures — the edge cases, the multiple-cause situations, the rare failure modes that don't show up clearly through a screen — the first ten minutes is not enough, and the right answer is to call a working cleaner who can spend the next two hours.
The tool's job is not to replace the cleaner. It is to handle the cases where the cleaner is not necessary, and to identify cleanly the cases where the cleaner is.
I keep an old textbook on my desk from the trade school I went through in Copenhagen. It is in Danish, it was written by the senior instructor's predecessor sometime in the 1970s, and the section on diagnosis is six pages long. The four questions are in there, on page 142, with a flow chart that traces the same branching pattern the Streak Diagnostic implements. The textbook has eleven verdicts to the diagnostic's seventeen — we've added a few that the original framework didn't account for (plasticizer migration, post-pressure-washing residue, certain fabricating-debris failure modes that weren't well-understood until the 1990s) — but the structure is the same. The four questions. The visual archetypes. The honest I cannot tell from here fallback when the cues don't match any known case.
I sometimes joke that the most modern thing about the Streak Diagnostic is that it doesn't require you to be in Copenhagen to use it. The framework itself is older than I am.
The right diagnosis is the foundation of the right fix. The right fix, applied to the wrong diagnosis, makes most cleaning problems worse. The Streak Diagnostic exists because the alternative — five years of trade-school training for every homeowner who wants to clean their own windows — is not realistic. Four questions on a screen is realistic. Four questions on a screen is what we have.
Abby Giordano trained as a production cleaner in Copenhagen before moving to New York in 2008. He has cleaned the glass on the Empire State Building lobby renovation, written for Pro Window Cleaner Magazine since 2013, and serves as Editor at Large at Window Washing Guide. He lives in Brooklyn, where he still runs a small route on Saturdays because, as he puts it, you forget the work if you stop doing it.
Abby trained as a production cleaner in Copenhagen before moving to New York in 2008. He has cleaned the glass on the Empire State Building lobby renovation, written for Pro Window Cleaner Magazine since 2013, and serves as Editor at Large at Window Washing Guide. He lives in Brooklyn, where he still runs a small route on Saturdays because, as he puts it, you forget the work if you stop doing it.