Diagnose a Coating Failure
Three questions, answered from memory on the jobsite. What the floor looks like, when the fault showed itself, and what the slab and the weather were doing. You get the candidates ranked, with the reason each one ranked where it did.
Answer as many of the three questions as you can and the tool ranks the pages that fit, with the reason each one ranked where it did. Not sure is a real answer on every question. If you already know what the defect is called, the troubleshooting index lists every page by name and will be quicker.
What it is likely to be
Answer at least one question above and the candidates appear here, ranked, each with the reason it ranked where it did.
What the ranking is doing, and what it is not
The three questions are not weighted equally, and the order matters more than the arithmetic. What the floor looks like counts for the most, because appearance is the strongest evidence anybody standing on it has. When the fault appeared counts for less than that and more than what follows. The conditions on the day count for the least, and they should: a damp slab makes several failures likelier at once without being evidence for any single one of them.
That ordering is the whole difference between a diagnosis and a survey. If ticking three conditions could outrank what you can plainly see, the tool would start recommending pages on circumstance alone, and it would look just as confident doing it.
What this cannot do is test anything. It has not seen your floor, it does not know which product went down, and a fault that is rare across every job can be the obvious answer on yours. Treat the ranking as a shortlist rather than a verdict. Every page it sends you to carries the physical check that settles the question, and that check is the thing to act on. Where the slab is implicated, measure it: a moisture reading costs very little next to coating a floor twice.
Why this is a different page from the visual identifier
The visual identifier sorts the same library by appearance alone, and when you can see which group a floor belongs in, it is the faster route. It works as a filter: find the group, read the tell, go.
Some faults do not separate by sight. A bubble and an osmotic blister sit in the same group, look the same, and share their tell almost word for word, so the visual route can only tell you to cut one open and see whether air or liquid comes out. Asking when it appeared separates them without touching the floor, because one is born in the wet film and the other turns up months after handover. The conditions question then says which is likelier on your slab.
The other difference is who each page serves. The visual identifier needs you to classify what you are looking at before it is any use. This one accepts not sure on every question and still ranks on what is left, which is the reader who has a floor, a problem, and no word for either.
Write the conditions down before they are gone
The second and third questions on this page are both much easier to answer honestly from a note taken on the day than from memory a fortnight later, and conditions are the one piece of evidence that disappears completely if nobody records them. Note when the coat went down, the air and slab temperature and the relative humidity at the time, the product and batch, the mix ratio and pot life, and how long the coat sat before the next one went on. Photograph the fault with something in frame for scale.
Where more than one fault is present, and that is common rather than unusual, work in the order things have to be fixed. Anything to do with the slab, meaning moisture, contamination and preparation, comes before anything to do with the film. A perfect topcoat over an unresolved slab problem fails again on the same schedule the first one did.
Frequently Asked Questions
How is this different from identifying the defect by sight?
The visual identifier asks one question and this asks three, and the extra two do work the first one cannot.
Sorting by appearance is a filter. It puts a floor into a group and hands you a physical test to separate the members of that group. That is the right tool when you can see the difference, and it is quicker.
Some defects look the same, though. A bubble and an osmotic blister are both closed domes in the same family, and the only visual test that separates them is cutting one open to see whether air or liquid comes out. When it appeared separates them without touching the floor, because a bubble is born in the wet film and a blister turns up months after the job was signed off. The conditions then say which is likelier on your slab. That is what the extra two questions buy you.
How does it decide the order?
By weighing the three answers against each other, and the weighting is deliberate rather than arbitrary.
What you can see counts for the most, because appearance is the strongest evidence you have. When it appeared counts for less than that but more than the rest. The job conditions count for the least, and they should: a damp slab makes several faults likelier without being evidence for any one of them in particular.
Every candidate shows the reasons it scored, so you can see the working rather than taking the order on trust. If a result looks wrong to you, the reasons will usually show you which answer moved it.
What if I do not know the answer to one of the questions?
Say so. Every question has a "not sure" option and it is a real answer rather than a way of skipping ahead.
An unanswered axis simply contributes nothing to the ranking, and the tool ranks on whatever is left. Answering one question of three still narrows twenty pages to a handful, which is the point: the reader this exists for is standing on a floor with no vocabulary for what is wrong with it, and demanding three confident answers before helping would fail exactly that person.
Can the top answer be wrong?
Yes, and it is worth being clear about that before you act on it.
This ranks likelihood from three answers. It does not test anything, it cannot see your floor, and it does not know what product went down. A fault that is rare in general can be the obvious answer on a particular job, and nothing here would know.
Treat the ranking as a shortlist rather than a verdict. Every page it sends you to carries the physical check that actually settles the question, and that check is what you should act on. Where the slab is implicated, test it: a moisture reading is cheap next to coating a floor twice.
Why does it sometimes list a lot of candidates?
Because a short answer set genuinely does not narrow things much, and the tool would rather show you that than hide it.
Nothing is capped and nothing is trimmed to a tidy five. If eleven pages match something you said, eleven is the honest number, and the ones near the bottom are there because they matched a single answer. Adding a second or third answer collapses the list quickly.
The candidates are split into the strongest matches and the rest, so a long tail does not bury the answer, but the tail is always counted and always reachable.
What should I write down before I start?
The conditions and the material, because both get lost within a day and both decide whether a repair holds.
Note when the coat went down, the air and slab temperature and the relative humidity at the time, which product and batch it was, the mix ratio, and how long it sat before the next coat. Photograph the fault with something in frame for scale.
The second question on this page asks when the fault appeared, and the third asks what the conditions were. Both are much easier to answer honestly from a note taken on the day than from memory a fortnight later.