How to Level a Subfloor Before Installing New Flooring

Most pages about subfloor leveling are published by somebody selling a bag of leveler or an install crew. So they’re vague about the one thing that actually matters, which is the number your specific floor has to hit. We don’t sell either. Here are the real specs, the places the industry contradicts itself, and the point where you should put the tools down.

Level and flat are not the same thing

This trips up almost everybody, including plenty of contractors. Level means horizontal, the bubble sitting dead center. Flat means no humps and no dips, whether or not the whole floor tilts.

Your floor needs to be flat. It does not need to be level. Read every manufacturer spec you can find and they’re all written as a deviation under a straightedge, never as a bubble reading. The tile trade has a saying for it: level is desired, flat is required.

This matters because a gently sloped but perfectly flat room is usually fine for the floor, while a floor with a hump in the middle is not, even if a level says it is perfect. An old house that leans a little across the living room can generally take laminate without complaint. Chasing level when you needed flat is how people waste a weekend and a hundred dollars of leveler.

One caveat, because the brands split here too. A gentle settle is one thing, a genuine slope is another, and several laminate makers flatly refuse to warranty a floating floor laid on a ramp or a sloped surface. If your room is noticeably pitched rather than just a bit tired, check your own sheet before you decide the slope is somebody else’s problem.

How flat does your subfloor actually need to be?

The number depends on what’s going on top, and it is tighter than people expect. Here is what the trade bodies publish:

• Nail-down solid hardwood, fasteners 1 1/2 inches or longer: 1/4 inch across 10 feet, or 3/16 across 6 (NWFA)
• Glue-down, floating, or short-fastener wood: 3/16 inch across 10 feet, or 1/8 across 6 (NWFA)
• Sheet vinyl and other resilient floors: 3/16 inch across a 10 foot span. This is the figure the resilient trade works to and it gets cited to ASTM F710 everywhere, but F710 itself sits behind a paywall, so we will not pretend we have read the clause.
• Ceramic and porcelain tile, every edge under 15 inches: 1/4 inch across 10 feet (ANSI A108.02)
• Large format tile, any edge of 15 inches or more: 1/8 inch across 10 feet

Notice that nailed hardwood gets the loosest tolerance of the lot. That isn’t an accident. The nails drag every board flat against whatever it lands on. A floating floor has nothing dragging it anywhere, so it copies the subfloor exactly, and the click joints eat the difference until they give up.

The laminate number everyone quotes, and where it comes from

You will read that the laminate association requires 3/16 inch over 10 feet. It sets no such standard. Its own installation guidance says the subfloor must be flat, and then defines flat as no ridges or low spots beyond what the manufacturer’s instructions allow. That is the whole of it. The association hands the question straight back to whoever made your planks.

What it does publish is more useful anyway: a table of what each member brand demands, and the brands do not agree. The tightest asks for 1/16 inch over 3 feet. The loosest allows 3/8 inch over 10 feet, which is double what most of the industry accepts. In between, most of them land on 3/16 inch over 10 feet, which is exactly why that figure got mistaken for a standard.

So 3/16 inch over 10 feet is real, sensible, and the most common answer. It is not an association rule, no matter how many pages say so, and if you happen to have bought the loosest brand or the strictest one it is not your answer at all. Go read the sheet in your box. Whatever it prints is what a warranty claim gets measured against.

How to check if a floor is level

Lay a 10 foot straightedge across the floor and slide a tape or a set of feeler gauges under it, hunting for daylight. Do it in a grid, both directions, and mark every high spot and every low spot right on the subfloor with a pencil. That map is the whole job.

One thing to know before you go buy a tool. A 4 foot level will lie to you here. Rest it on top of a hump and it rocks like a seesaw, and both ends read low, so a hump reads as two dips. The spec is written over 10 feet because that’s the span that tells the truth. Borrow a long one.

A laser level is handy for spotting where the trouble lives across a big room. It doesn’t replace the straightedge. The straightedge is what the spec is written against.

Fixing high spots

On concrete you grind. On a wood subfloor you sand or plane. Both are loud, dusty, and satisfying.

Here is the thing nobody tells you: grinding cannot fill a low spot. Grinding only removes material, so all you can do is chew the hump down. If you attack the ground around a dip you just make a wider, shallower dip and move the problem. Low spots get filled. High spots get ground. Do them in that order and check with the straightedge between passes.

The rule itself binds employers, not you in your own basement. Two things follow from that, and people only ever remember the first. Silicosis does not care whose house it is. And the rule does bind the crew you hire, so you are entitled to see those controls on your job.

Mixing a bag of leveler gets treated as the harmless part, and we used to say so ourselves. That was wrong, and the bag disagrees with us. Dry leveler powder contains crystalline silica, and the safety sheet on it carries the word Danger and the sentence that it may cause cancer by inhalation and damages lungs through repeated exposure. That is the manufacturer talking about its own product, not us being dramatic.

Mixing is not on OSHA’s list of tasks with a prescribed control recipe, and it is tempting to read that absence as permission. It means the opposite. Tasks off the list get no safe harbor and fall under the exposure limit itself, and OSHA has said in writing that dumping bags into a mixer over a stretch of time can foreseeably push exposure past the level where something has to be done about it. Ardex is blunter still: its own sheet puts a HEPA vacuum on the mixing step, under a heading about OSHA compliance. Custom tells you to wear a NIOSH approved respirator and not to breathe the dust.

So do what the bag says. Cut it low, pour it slowly so you never raise a plume, and wear a real respirator, an N95 at the very least. And notice the fine print on the one reassuring line in the whole safety sheet, which says mixing to the directions produces no measurable airborne silica. The directions are the dust controls. The reassurance is conditional on doing the thing we used to tell you that you could skip.

Subfloor leveling: patch, self-leveler, or neither

Three products get confused constantly. Floor patch is a stiff, trowel-applied compound for small dips, gouges, and seams. Self-leveling underlayment is a liquid you pour across a large area. A mortar bed is a thick sand and cement bed, a specialty tile technique.

And thinset is none of them. Mapei, who sells all of this, puts it about as bluntly as a manufacturer ever does, saying in its own words that mortars cannot be used to level or patch a floor, ever, exclamation mark and all. The reason is that a mortar goes unstable once you push it past its intended thickness, and most thinsets are only good to a quarter of an inch. If somebody suggests troweling thinset to smooth your floor out, they are guessing.

Self-leveler does not actually level itself

The name oversells it. The stuff flows and finds its own plane, but only for about ten minutes, and only across a limited distance before it starts to stiffen. Real installers move it with a gauge rake and work fast. Pour a big room by yourself and read the bag afterward and you will understand why people mix in teams.

Primer is not optional

Skipping the primer is the classic way a self-leveler fails, and TEC says plainly in its own guidance that failing to use the recommended primer may result in installation failure. Nobody publishes a ranking of failure causes, so treat that as the well-known mistake rather than as a statistic.

The primer does two jobs, and the first is the one people miss. It stops a porous subfloor from drinking the water back out of the mix, and that water is what the leveler needs to cure properly. It is also a bonding agent, which we used to deny. What you get without it is a hard-looking floor that quietly delaminates and crunches underfoot a year later.

Depth, strength, and the wait

Pour depth is where you must read your own bag, because products differ by a factor of four. Custom’s LevelQuik RS pours from a featheredge to about an inch and a half. Ardex K 15 goes to an inch and a half on its own, or five inches cut with pea gravel. Mapei’s deep-pour Novoplan DPL takes four inches in a single lift.

We used to say we had found two copies of one data sheet giving different maximums, without naming it, which is not much use to you. It is Custom Building Products’ LevelQuik RS, and it is worse than we made it sound. The older sheet contradicts itself inside one document: the description says an inch and a half, and the benefits and application sections both say one inch. The current sheet, same product and same document number, settles on an inch and a half throughout. So make sure the sheet you are reading is the current one, and if a number matters, get it from the bag in your hands.

Strength has a floor, and the floor is set by the flooring, not by the leveler. Shaw asks that any patching or leveling compound under its planks reach at least 3,000 psi. Mohawk asks the same. Armstrong is stricter and wants 3,500. So 3,000 is the number to check against, and it is a minimum you accept rather than a target you chase.

Real products clear it comfortably and they spread out more than we used to say. We previously told you most decent levelers land between 4,300 and 5,500 psi, which quietly wrote off several good ones. The honest spread runs from roughly 3,500 to 6,000: Mapei’s deep-pour sits near 3,500, its Novoplan 2 Plus around 4,200, Custom’s LevelQuik above 4,300, Ardex K 15 at 5,500, and TEC’s fiber-reinforced at 6,000. Every one of those passes the 3,000 psi bar. Do not use a narrow band as a screening test and reject a product that was fine.

Then wait, and read the wait time for the covering you are actually laying, because this is where a good pour gets ruined. Ceramic, porcelain, and quarry tile can go down fast, sometimes in as little as four to six hours, because they do not mind a bit of residual moisture. Ardex K 15 allows six. Custom’s LevelQuik RS allows four. Mapei’s Novoplan 2 Plus still wants a full 24, so fast is a property of the product and not of tile.

Natural stone is not in that fast group, and we used to say it was. We were mistaken, in the way that ruins a floor. Marble, limestone, and the rest are moisture-sensitive, and they stain, spall, and curl over a damp substrate. Every fast figure in every sheet we read is scoped to moisture-insensitive tile by name, and stone is never on that list. Stone belongs with wood, laminate, and vinyl: 14 to 24 hours at a minimum, and Mapei wants two to three days at a one inch pour.

Can you pour it straight onto plywood?

Sometimes, and it is genuinely product by product rather than chemistry by chemistry. Custom’s LevelQuik RS wants metal lath over a wood subfloor, 2.5 pounds per square yard, fastened every six to eight inches. The fiber-reinforced formulas are built to go without it: TEC’s fiber-reinforced underlayment, Laticrete’s NXT Level Plus, and CustomTech’s TechLevel WSF all say so, though each still ties a minimum thickness to your joist spacing. Mapei’s Novoplan 2 Plus will not go over a plank subfloor at all until you have glued and screwed down at least one layer of 5/8 inch plywood. Anybody who tells you to just prime it and pour has not read the sheet for the product in your cart.

Prepping a concrete subfloor

Concrete brings its own problem, and it isn’t flatness. It’s water. A slab pushes moisture upward for years, and that vapor destroys adhesives, cups planks, and grows things you don’t want under your floor.

How to test concrete slab for moisture

Two standardized tests exist. One of them, ASTM F1869, parks a dish of dry calcium chloride on the slab, leaves it sealed under a dome for 60 to 72 hours, and weighs the water it captures, scored in pounds per 1,000 square feet over 24 hours. The other, ASTM F2170, sinks a probe to 40% of the slab’s depth and reads the humidity from inside the concrete itself.

Most flooring makers now ask for the probe. The argument against the dish is that it only reads near the surface and misses what lies deeper, which is fair, though notice who makes that argument loudest: the companies selling probes. Take it as a real point with a commercial edge on it, rather than as something the standard itself says.

If you go looking, you will find page after page saying the probe has to sit for 72 hours before you read it. That was true, and it stopped being true in 2018, when the standard cut the wait to a minimum of 24. The internet has not caught up. Yours needs a day, not three.

The passing number is not universal, which is the part that catches people. Bruce wants 3 pounds or under, or 75% RH. Mohawk’s RevWood installation instructions permit as much as 5 pounds, or under 80% RH. The wood flooring association’s own baseline says anything over 3 pounds should wait for the slab to dry or get a Class I vapor retarder. Same slab, three different verdicts. Go by the product you’re installing.

Cured and dry are two different clocks

Consumer articles collapse these into one number and it causes real damage. Curing is not the slab getting strong, which is how we used to put it and how nearly everyone puts it. Curing is the deliberate act of holding moisture and temperature in the fresh concrete so the cement can hydrate properly. Strength gain is what that buys you, and it does not stop at 28 days either; that date is a design benchmark, not a finish line. Drying is a different thing entirely: it is moisture leaving the slab, and it takes vastly longer than either.

You will hear roughly 30 days of drying per inch of thickness. Treat that as trade lore and nothing more, because no engineering body publishes it and the real world does not obey it. Measured work has put an ordinary 4 inch slab at around 46 days to hit the usual threshold, and a lightweight slab at 183. Either way, an ordinary residential slab is looking at months, not weeks.

The trade bodies land in different places, which is worth seeing rather than being told to trust one. The laminate association asks for 90 days minimum under laminate, plus a 6 mil vapor barrier, though its own older guide says 60, so it does not entirely agree with itself. The tile association names 14 to 28 days as the usual window and says plainly that no single number can be stated for everyone. Neither one is a substitute for a reading. A slab can be plenty strong and still far too wet, and only a test knows which yours is.

Plywood vs OSB subfloor

Code treats them as structural equals and you can lay a good floor over either. They fail differently when they get wet, and that difference lands right where you care about it.

Plywood takes on water fast and gives it back fast, and when it’s had too much it delaminates into layers. OSB swells at the panel edges, and here’s the catch: much of that swelling never comes back out. Some of it does, which is why the panel makers tell you to let a wet subfloor dry before you reach for a sander, since you may be sanding a ridge that was going to shrink on its own. What is left after it dries is a ridge running along every panel seam, and that is precisely where your straightedge is going to find a hump. Sand those flat if the panel is otherwise sound. Replace it if it’s soft.

A spongy spot underfoot means the panel has already come apart internally. Sanding won’t save that.

Squeaks and loose panels

Fix the squeaks now. Once the new floor is down, every one of them is permanent, and every one of them will annoy you for a decade.

Subfloors squeak because nails work their way back out of the joists, and the usual culprit is not the seasons but the framing lumber drying out after it was installed green, popping the nails as it shrinks. Unsupported panel edges flexing does the rest. The trade answer is glue and screw, and the panel association recommends a glued floor system outright. Find the joists, run a bead of construction adhesive where you can reach, and drive screws down into them.

The fastening schedule from the residential code is 6 inches on center along panel edges and 12 inches through the field. Leave the 1/8 inch gap between panels that the panel makers ask for. It’s there so the subfloor itself can expand without buckling, which is the same physics as the expansion gap in your finished floor.

How to fix a damaged subfloor panel

Rot, a long-running leak, or mold means the panel comes out. Cut it back to the middle of the joists on either side so the new piece has something to land on, sister a piece of lumber alongside if the joist itself is soft, and screw the replacement down on the same schedule.

Do this before anything else on this page. There is no leveling compound that fixes rot, and pouring one over a wet panel seals the water in.

When to put the tools down

• The floor bounces or deflects underfoot. The old rule of thumb for tile is that the floor should move no more than its span divided by 360, and the stone institute asks for twice that stiffness, L over 720, for stone over a wood frame. Worth knowing that the tile association itself dropped the deflection rule years ago, after seeing tile fail on floors stiffer than the number said was fine. Either way this is a framing question, not a flooring one.
• The slab fails its moisture test. Waiting and a mitigation system are the only real answers, and neither is a bag from the aisle.
• The floor slopes noticeably and the slope seems to be getting worse, or it comes with new cracks in drywall and doors that stopped latching. That is a foundation conversation, and a leveler will hide it, not fix it.
• Joists are sagging, undersized, or rotten across a broad area.

On that third one, we deliberately aren’t giving you an inch-per-foot threshold, and the reason deserves saying plainly. Engineers do have criteria for this. There are published tolerances for how level a slab should be poured, and there is a whole body of work on how much differential movement a foundation can take before things crack. What none of that is, is a pass or fail number a homeowner can hold a tape measure against. Those tidy inch-per-foot figures you find online come almost entirely from companies that sell foundation repair, which is a poor place to learn whether you need foundation repair. If the floor is moving, get an engineer, not a fraction from a blog.