Why Is My Epoxy or Flake Floor Peeling?
Peeling (delamination) almost always means the bond never really formed, and the cause is usually surface prep rather than the coating itself. Here’s what causes it and how to prevent it on the next job.
Last updated August 21, 2026
Peeling (delamination) almost always means the bond to the concrete never really formed. The single biggest cause by far is surface prep: a smooth or laitance-covered slab gives the coating nothing to mechanically key into, no matter how clean it looks. Moisture, contamination, and driving on an undercured coating account for most of the rest. Grinding or shot-blasting to the right profile prevents the large majority of it.
Most likely causes
Poor surface preparation (no mechanical profile)
- Concrete has a weak surface layer called laitance that has to be ground or shot-blasted off, and acid etching or simply cleaning the slab isn’t enough on its own
- A surface that’s too smooth gives the coating nothing to grip, even when it looks perfectly clean
- Fix: grind or shot-blast to the coating manufacturer’s specified concrete surface profile (CSP) before priming
Moisture vapor from an untested slab
- Vapor pressure building up under the coating from an unsealed or under-tested slab eventually pushes the film off from underneath
- This is the same root cause behind bubbling, and peeling is often just further along
- Fix: test RH or MVER before coating, and apply a moisture vapor barrier if levels exceed the coating’s rated limit
Contamination or coating over unremoved blush
- Oil, grease, dust, old sealers, and curing compounds left on the slab block the bond just as effectively as a bad profile
- Recoating over amine blush that was never removed is one of the most common causes of a topcoat peeling later
- Fix: confirm the slab is free of contamination before priming, and always remove blush before the next coat
Traffic before the coating fully cures
- “Walk-on” time and rated vehicle-traffic cure time are very different numbers, and driving on it early, especially with hot tires, can lift a coating that hasn’t fully hardened
- Off-ratio mixing (too little hardener) can leave a coating permanently undercured, with the same result even after waiting
- Fix: respect the product’s rated vehicle-traffic cure time, and mix to the exact Part A:B ratio on the TDS
How to tell which failure you are looking at
All four causes above end the same way, with coating coming off the floor, so the fix depends entirely on telling them apart before you strip anything. Turn a lifted piece over and look at its underside. That one surface answers most of the question.
Grey concrete stuck to the back
- The bond held. The coating pulled the top of the slab off with it, which means the concrete itself was weak
- Usually laitance that was never ground away, or a slab that was overworked or over watered when it was poured
- Points to: surface preparation, and sometimes a slab that needs more removal than a light grind
Clean, bare concrete underneath
- The coating released cleanly from a sound slab, so the concrete was strong and the bond simply never formed
- Look for a glossy or powdery slab surface, a sealer sheen, or an oily patch that repels water
- Points to: contamination, an unremoved sealer or curing compound, or a profile that was far too smooth
Damp, chalky or white residue
- Moisture has been moving through the slab and depositing salts under the film
- Often worst at slab edges, in low spots, or over an area with no vapor barrier beneath it, and it tends to return in the same places after a recoat
- Points to: vapor pressure, the same root cause as bubbling, taken further
It is one coat leaving another
- The concrete is not involved at all. The topcoat is releasing from the base coat below it
- Check whether the failure follows the recoat, and whether the surface underneath is cloudy or waxy
- Points to: intercoat adhesion, usually a missed recoat window or coating over unremoved amine blush
Get the surface prep time right
Surface prep is the single biggest cause of peeling. Estimate the grinding time to build the profile the coating needs before you ever open a kit.
Full result gives total grinding hours and days to complete the job.
Open the live calculator →What the preparation actually has to achieve
Prep is not cleaning. It has two separate jobs, and passing one of them while failing the other still leaves you with a floor that peels.
Remove the weak layer
- The top of a power troweled slab is laitance: fine cement and water that rose during finishing and set into a layer far weaker than the concrete beneath it
- A coating bonded to laitance is only ever as strong as the laitance, so it can pass every visual check and still lift later with grey dust on its back
Open a profile
- Concrete surface profile (CSP) is the ICRI scale from 1 to 10 that describes how open the surface is. Most epoxy and polyaspartic systems call for CSP 2 to 3
- Diamond grinding or shot blasting produce that consistently. Acid etching does not: it can leave the laitance in place, and it leaves behind residue that has to be neutralised and rinsed
Prove it before you coat
- Drop water on the prepared slab. It should darken and soak in within seconds. If it beads, something is still sealing the surface and the coating will sit on that instead of the concrete
- Where the job warrants it, a pull off test to the coating manufacturer’s stated value is the only direct measure of what the bond will actually do
Worked example: the prep against the redo
Prep is the line contractors cut when a quote is tight, so it is worth putting hours against it. Take a 540 ft² single garage at the calculator’s default 300 ft²/hr, which is a rate for one pass, not for the whole job. Two passes to remove laitance and open the profile: 540 ÷ 300 = 1.8 hours per pass, so 3.6 hours of grinding in total, comfortably inside a single day.
Now price the failure. A floor that has delaminated has to have the old coating ground off before anything else happens, and that removal buys you nothing: at the end of it you are standing on a bare slab that still needs the same two passes you skipped. The cheapest possible redo is therefore 7.2 hours of grinding rather than 3.6, and that is before you have bought a second set of materials or gone back to the site. On a 1,500 ft² commercial floor the same two passes run 5 hours each, so 10 hours of prep turns into 20.
That is the whole economic case for prep, and it does not depend on any market rate: the second grind costs whatever the first one costs, and you pay for it twice. What concrete grinding costs per sq ft puts money against those hours.
How to fix it if it’s already happened
- A small, isolated patch can often be ground out and repaired locally
- Widespread peeling usually means stripping or grinding off the whole coating and reapplying, because patching over a bond failure just moves the problem
- Fix the root cause before recoating, whether that’s testing moisture, improving the profile, or removing contamination. Otherwise the same failure happens again
Common mistakes
- Treating a clean slab as a prepared slab. Sweeping, degreasing and vacuuming all matter, and none of them remove laitance or open a profile
- Etching because grinding was not available, then coating anyway. Etching alone is inconsistent on a hard troweled slab and leaves residue that has to be neutralised and rinsed
- Recoating a failed area without turning a piece over first. The underside tells you whether you are fixing prep, moisture, contamination or intercoat adhesion, and those have four different fixes
- Reading the walk on time as the drive on time. They can differ by days, and hot tires on a coating that has not reached its rated vehicle cure will lift it
- Grinding a floor that peeled from moisture and coating it again unchanged. Nothing about the vapor pressure has moved, so the same failure is already scheduled
Frequently Asked Questions
Is all peeling caused by bad surface prep?
No. Moisture, contamination (including unremoved amine blush), and driving on an undercured coating can all cause it too. But surface prep is by far the most common cause, responsible for the large majority of adhesion failures.
Can a peeling floor be patched, or does it need to be redone?
It depends how much has delaminated. A small, isolated patch can sometimes be ground out and repaired locally. Widespread peeling usually means the whole coating needs to be stripped or ground off and reapplied. Patching over a bond failure just moves the problem somewhere else.
How long should I wait before parking a vehicle on a new coating?
Follow the product’s rated vehicle-traffic cure time on its TDS, not the walk-on time, because the two can differ by days. Driving on it too soon, especially with hot tires, is a common cause of localized peeling.
Does acid etching count as surface prep?
Not reliably on its own for most coating systems. Diamond grinding or shot blasting is what consistently removes the weak surface layer (laitance) and creates a mechanical profile for the coating to grip. Acid etching alone is inconsistent and often insufficient.
How can I tell what caused it before I strip anything?
Turn a lifted piece over and look at the underside. Grey concrete stuck to the back means the bond held and the slab gave way, which points at laitance that was never ground off. Clean bare concrete means the slab was sound and the bond never formed, which points at contamination or too smooth a profile. A damp or chalky residue points at moisture moving through the slab. And if the coating is leaving another coat rather than the concrete, it is an intercoat problem, usually a missed recoat window or unremoved amine blush. Those four have four different fixes, so it is worth the thirty seconds.
Can I check the preparation is good before I coat over it?
Yes, and it takes a minute. Drop water on the prepared slab: it should darken and soak in within a few seconds. If it beads or sits on the surface, something is still sealing the concrete and the coating will bond to that rather than to the slab. On jobs where a failure would be expensive, a pull off test against the value in the coating manufacturer's data sheet is the only direct measurement of what the bond will actually do.