How do I know whether a metallic floor is actually defective?

By comparing it to the sample board, and by very little else. This is the uncomfortable part of metallic work and it is worth saying plainly. A solid color floor has an objective standard: it should be one shade, and any departure from that is a defect you can point at. A metallic floor has no such standard. The pattern is generated by pigment moving through wet resin, and no two pours are the same, so "it does not look like I expected" is not by itself a fault. What gives you a standard is a sample board poured with the same products, in the same sequence, by the same technique, and approved by the client before the floor went down. That board is the specification. If the floor sits within the range of what the board shows, it is the product working. If the floor is markedly flatter, muddier, or more broken than the board, something in the application changed. Without a board you are arguing taste against taste, and the client usually wins that argument. Pour one on every metallic job, and keep it.

The floor came out flat with no depth. What causes washout?

Four causes, and the first two are the common ones. The metallic coat was spread too thin. A metallic coat is a build coat, not a color coat: the pigment needs a depth of resin to migrate through, and the illusion of depth comes from looking down into a translucent layer. Stretch that coat to cover more ground and you take away the thing the effect is made of. It covers far less area per gallon than a base coat for exactly this reason, and treating that number as generous is how the coat gets stretched. The base coat gives no contrast. A large part of what reads as depth is the base showing through the translucent metallic layer. A base coat close in value to the metallic leaves nothing for the pattern to read against, and the floor looks pale and vague even when the metallic coat was applied correctly. It was over-worked. The pattern is made by letting the pigment separate, not by moving it. Every extra pass re-blends what the last one separated, so a crew that keeps working an area to improve it is homogenizing it. It was worked too late. Once the resin starts to gel the pigment can no longer migrate, and anything done after that point marks the surface without moving the color.

How do I tell an induced cell from a fisheye?

Four tests, and any one of them usually settles it. Location. Cells appear where solvent or heat was applied, so they follow the pattern of your own application. Fisheyes appear at random points that correspond to nothing you did. The rim. A cell opens as a smooth gradient, with the color thinning out toward the middle. A fisheye has a sharp raised ring where the coating pulled back from a point of low surface tension, and you can usually feel it. The center. A fisheye frequently has a visible speck at its middle, which is the contaminant that caused it. A cell has nothing at its center but the coat below. Repeatability. Contamination is in or on the substrate, so fisheyes come back in the same places when you recoat. Cells do not. This distinction decides the whole repair, which is why it is worth the two minutes. Fisheyes mean contamination, and the fix is decontamination and re-prep. Cells mean your solvent dose or your torch was heavier than intended, and the fix is a technique change on the next coat.

What makes the cells blow open into craters?

Almost always too much solvent, or solvent onto a coat that was too thin to survive it. Denatured alcohol or a similar solvent works by locally dropping the surface tension so the coating pulls back and reveals what is beneath. That is a controlled version of the same mechanism that produces a fisheye. Dose it heavily, spray it from too close, or let it pool, and the coating does not pull back into a cell, it gets displaced entirely and leaves a bare crater with a hard edge. A torch or heat gun does the same thing by a different route. Held too close or too long it drives the resin away from the spot rather than just releasing air from it. And both failures get much more likely on a thin metallic coat, which is why washout and craters so often turn up on the same floor. A coat with a proper build has enough material to flow back and close a cell into a soft edge. A stretched coat has nothing in reserve, so the same solvent dose that would have made a cell makes a hole.

The pattern is there but it looks muddy. What went wrong?

Mottling is usually a color choice rather than an application fault, which is why it is the hardest of the three to fix on site. Colors too close in value gray each other out. Metallic pigments are translucent and they layer, so two or three shades within a narrow value range blend into a single dull tone rather than reading as separate movements. Too many colors do the same thing faster. Each additional pigment gives the others something more to muddy against, and the practical limit is lower than most people expect. Uneven film build contributes. Thick and thin areas hold different amounts of pigment and move at different rates, so a floor laid with an inconsistent spread reads as blotchy even with well chosen colors. And working an area after the cells have set tears the pattern rather than refining it, leaving smears that look like dirt. The first two of those are decided before anyone opens a kit, which is the argument for the sample board again: a color combination that mottles will mottle on a two foot board just as reliably as on a warehouse.

Can I repair a bad patch without redoing the floor?

Less easily than on a solid color floor, and the intuition runs the wrong way here. The assumption people make is that because a metallic floor is random, a patch will disappear into it. The opposite is true. A metallic pattern is continuous across the pour, with movement that runs through the whole floor, and a patch is a self-contained island of pattern that stops at its own edge. The eye finds that boundary immediately, even when the colors match perfectly. What does work is another metallic coat over the whole area. Metallic layers are translucent, so a second pass adds depth to a washed-out floor rather than covering it, and it can rescue a pour that came out flat. It will not save a floor with open craters, because the craters will still telegraph through. Craters and heavy mottling generally mean grinding back and starting the metallic coat again. The base coat can often be kept if it is sound and correctly colored, which takes a good deal of cost out of the redo. Before any of that, check the recoat window. A metallic coat that has cured past its window needs abrading and carries an adhesion question on top of an appearance one.

Does the topcoat change how the effect looks?

Yes, and more than most people allow for, which makes it worth testing on the board rather than discovering on the floor. A clear topcoat wets out the metallic layer and generally deepens it, so a floor that looks slightly flat before sealing often improves. That is a real effect rather than wishful thinking, but it is a modest improvement rather than a rescue, and it will not manufacture a pattern that is not there. A satin or matte topcoat works the other way. Scattering light at the surface reduces the sense of looking down into the film, so the same floor reads flatter under a matte seal than under a gloss one. If depth is the point of the floor, that is a specification decision rather than a finishing detail. The manipulated metallic surface also tends to take more topcoat than a flat floor of the same area, because it is not perfectly level and the seal coat has more to fill. So seal the sample board with the same topcoat you intend to use on the floor. A board approved unsealed is not a board approved.

Is a failed metallic effect a performance problem?

Usually not, and separating the two matters because it changes what you owe the client. If the pigment moved wrongly but the resin and hardener were correctly proportioned and mixed, the film cures and performs exactly as it should. The floor is as durable and as chemically resistant as a floor with a perfect pattern. What has failed is appearance, and on a decorative floor appearance is what was bought, so that is not a small thing, but it is not a structural one either. There are two exceptions worth checking before you agree it is cosmetic. Craters that go through to the base coat are a thin spot in the system, and a thin spot in a floor that takes traffic or spills is a durability question rather than a looks one. And if pale or blotchy areas are also soft or tacky, then the mix was short of more than pigment, and you have a cure failure that happens to be showing as color. Press a fingernail into the worst areas. If everything is hard and the film is continuous, it is an appearance defect, and the conversation is about the sample board.

Metallic Effect Failures: When the Pattern Is Wrong

Every other coating defect starts from the same assumption: the floor should be uniform and it is not. A metallic floor breaks that, because the unevenness is the product you sold. This page is about telling which unevenness is the effect working and which is a fault, and the honest answer to that begins with a sample board rather than a description.

Quick answer

Put the floor next to the sample board before you call it a defect. A metallic floor is meant to be uneven, so unevenness is not the fault by itself. What you are looking for is variation that is too weak (washout), at the wrong scale (mottling), or torn open (craters). If the open marks appear where you never applied solvent and have a sharp raised rim, you have fisheyes instead, which is contamination and a completely different repair.

Which of the three you have

Flat and lifeless

The color is right but the floor has no depth and barely moves. It reads like a plain coat
Cause: a metallic coat spread too thin, a base with no contrast, or an area worked so long the pattern was blended back out
Usually uniform across the whole pour rather than in patches
This page. Often recoverable with another metallic coat over the top

Blotchy and muddy

There is movement, but it reads as dirt rather than flow. The colors have grayed each other out
Cause: usually the color selection rather than the application. Shades too close in value, or too many of them
Look for smearing where an area was worked after the cells had set
This page, though the honest fix starts at the sample board rather than on the floor

Open craters through to the base

Round bare areas with hard edges, showing the coat beneath
Cause: too much solvent, a torch held too close, or either of those onto a coat already too thin
If they sit where you never applied solvent, have a raised rim, or return in the same places on a recoat, go to fisheyes and craters

The problem with diagnosing a metallic floor

Every other defect page on this site can open the same way: the floor should be uniform, it is not, here is why. That sentence is doing more work than it looks like it is, because it supplies the standard. A solid color floor that varies is wrong by definition, and you can point at the variation.

A metallic floor has no such standard. The pattern comes from pigment migrating through wet resin while it cures, and no two pours are alike, so there is nothing to compare it to and no description precise enough to settle an argument. That is why the complaint on a metallic job is almost never that the floor is uneven. It is that the floor is not what the client pictured.

The sample board is the standard, and it is the only one available. Poured with the same products, in the same sequence, by the same technique, sealed with the same topcoat, and approved before the floor goes down. If the floor sits inside the range that board shows, it is the product working. If it is markedly flatter, muddier or more broken, something in the application changed and this page is about what. The metallic sizing guide already recommends a board for a different reason, which is that technique is not something to learn on a client's floor. This is the second reason, and on a disputed job it is the more valuable one.

Washout: four ways the depth disappears

The metallic coat was spread too thin. This is the first cause and by some distance. A metallic coat is a build coat rather than a color coat: the pigment needs a depth of resin to migrate through, and the depth you see is you looking down into a translucent layer. Stretch that coat to cover more ground and you have removed the thing the effect is made from. It covers far less area per gallon than the base coat underneath it, and reading that as slack in the estimate is exactly how it gets stretched
The base coat gives it nothing to read against. Much of what looks like depth is the base showing through. A base close in value to the metallic leaves the pattern with no contrast, and the floor comes out pale and vague even when the metallic coat itself was applied correctly
It was over-worked. The pattern is made by letting the pigment separate, not by moving it around. Every additional pass re-blends what the previous one separated, so a crew working an area harder to improve it is homogenizing it instead
It was worked too late. Once the resin begins to gel the pigment cannot migrate, and anything done after that marks the surface without moving any color

The first of those is the one with a tool attached. The metallic epoxy calculator sizes all three layers from one floor measurement so the metallic coat is ordered at its own rate rather than the base coat's, and the film thickness calculator will tell you what spread rate a target build actually demands. Use the rate on your own product's data sheet for both: metallic coats vary more between manufacturers than solid color coats do, so a generic number is a worse guide here than almost anywhere else on a job.

A cell or a fisheye? Four tests

This is the one distinction on the page worth stopping for, because the two look similar and the repairs have nothing in common. An induced cell and a fisheye are the same physics: a local drop in surface tension makes the coating pull back and expose what is under it. The difference is entirely whether you caused it on purpose.

Induced cells

Location: where solvent or heat was applied, following the pattern of your own work
Rim: a smooth gradient, color thinning toward the middle
Center: nothing but the coat below
On recoat: they do not come back in the same places
Fix: a lighter solvent dose or less heat next coat

Fisheyes

Location: random points corresponding to nothing you did
Rim: a sharp raised ring you can usually feel
Center: frequently a visible speck, which is the contaminant
On recoat: they return in the same places, because the cause is in the substrate
Fix: decontamination and re-prep. See fisheyes and craters

Getting this wrong is expensive in one direction and embarrassing in the other. Grinding and decontaminating a floor whose cells were induced deliberately buys nothing. Recoating over genuine contamination without prepping it reproduces the fisheyes in the same spots and proves the point at your own cost.

Why cells blow open into craters

Denatured alcohol works by dropping the surface tension locally so the coating retreats and reveals the layer beneath. Dose it heavily, spray from too close, or let it pool, and the coating does not retreat into a cell. It is displaced completely, leaving a bare crater with a hard edge. A torch does the same thing by a different route: held too close or too long it drives resin away from the spot rather than releasing air from it.

Both get far more likely on a thin coat, which is why washout and craters so often arrive on the same floor and get reported as two problems. A coat at proper build has material in reserve to flow back and close a cell into a soft edge. A stretched coat has none, so the same solvent dose that would have made a cell makes a hole instead. Fixing the build usually fixes both.

Mottling is decided before the kit is opened

Of the three failures this is the one least amenable to a technique change on site, because its main causes are choices made at quoting.

Colors too close in value gray each other out. Metallic pigments are translucent and they layer, so shades within a narrow value range blend into one dull tone instead of reading as separate movements
Too many colors do it faster. Every additional pigment gives the others more to muddy against, and the practical limit is lower than most people expect
Uneven film build. Thick and thin areas hold different amounts of pigment and move at different rates, so an inconsistent spread reads as blotchy even with well chosen colors
Working an area after the cells have set, which tears the pattern rather than refining it and leaves smears that genuinely do look like dirt

A color combination that mottles will mottle on a two foot board exactly as reliably as it does on a warehouse, at a thousandth of the cost. That is the whole argument for pouring one.

What can actually be repaired

The intuition here runs backwards, so it is worth stating the counter-intuitive part first.

A patch does not disappear into a metallic floor. People assume that because the pattern is random, a repair will blend. The opposite happens: a metallic pattern is continuous across the pour, and a patch is a self-contained island of pattern that stops at its own edge. The eye finds that boundary immediately even when the colors match perfectly
Another full metallic coat is the repair that works for washout. Metallic layers are translucent, so a second pass over the whole area adds depth rather than covering it, and it can rescue a pour that came out flat
Craters will telegraph through a recoat, so a second coat does not save a cratered floor. That one generally means grinding back and re-doing the metallic coat, though a sound and correctly colored base coat can usually be kept, which takes real cost out of the redo

One ordering matters, and it is the same one that governs every recoat on this site. A metallic coat that has cured past its recoat window needs abrading and carries an adhesion question on top of an appearance one. Settle that with intercoat delamination before covering anything, because a coat that restores the depth and then lets go has fixed nothing.

Is it cosmetic?

Usually, and saying so precisely matters because it decides what you owe the client. If the pigment moved wrongly but the resin and hardener were correctly proportioned and mixed, the film cures and performs exactly as intended. It is as durable and as chemically resistant as a floor with a perfect pattern. What failed is appearance, and on a decorative floor appearance is what was bought, so this is not a small thing. It is just not a structural one.

Two exceptions are worth ruling out before you agree it is looks only. Craters that reach the base coat are a thin spot in the system, and a thin spot in a floor taking traffic or spills is a durability question. And if the pale or blotchy areas are also soft or tacky, the mix was short of more than pigment, which makes it a cure failure showing itself as color.

On a floor that was meant to be one solid shade rather than a metallic one, the uneven color is a different fault with a different cause. That is streaking in solid color. And if you are still not sure which of these you are looking at, start from the symptom and work back with the defect router.