How to check a floor is level, and whether that is even the question.

Level and flat are different faults.

These two words get used interchangeably on site and they mean quite different things.

Level is a relationship with gravity. A level floor is horizontal; a floor that is out of level runs downhill in some direction. It is measured in degrees, or more usefully for floors, in millimetres of fall per metre.

Flat is a relationship with a plane, and gravity does not come into it at all. A flat floor has no humps or hollows: lay a straightedge anywhere on it, in any direction, and the straightedge touches along its whole length. It is measured as the largest gap under a straightedge of a stated length.

The two are genuinely independent. A concrete ramp is perfectly flat and not remotely level. A bowl-shaped slab that has settled in the middle can average out to dead level and still be hopeless to tile. When someone says "my floor isn't level" and means their laminate is bouncing in the middle of the room, they have a flatness problem, and pouring a compound to make the floor level is not necessarily going to fix it.

Measure both. They take the same twenty minutes and they lead to different work.

Which one your floor covering cares about.

This is the part that decides how much effort the survey deserves.

A floating floor does not care that the room slopes 15 mm end to end, because the whole assembly simply sits on the slope and nobody perceives it. What it cannot survive is a 6 mm hollow under a joint: the planks bridge the dip, flex every time you walk on them, and the click joint works itself apart. That is what a bouncing, creaking, gapping laminate floor actually is.

Tiles behave the same way but fail differently. A hump under large-format tiles produces lippage, where one tile edge stands proud of its neighbour, and there is no grouting your way out of it.

A wet room is the exception that proves the point: there the specification is a deliberate fall toward the drain, typically 10 to 20 mm per metre, and a floor that came out level is the defect. See the slope and gradient reference for the common figures.

The tolerances you are measuring against.

Flatness is always quoted as a deviation under a straightedge of a given length, and the length is half the specification. "3 mm" means nothing on its own.

In the UK, BS 8204 defines three surface regularity classes, all measured with a 2 m straightedge:

In North America the same idea is usually written as 3/16 inch in 10 feet, or 1/8 inch in 6 feet, which works out very close to SR2.

Most laminate and LVT manufacturers publish their own figure, and it is commonly 3 mm under a 2 m straightedge — that is, SR1. This catches people out: a floor that a builder would sign off as perfectly acceptable can be outside the flooring warranty. Look up the actual number for the product you have bought before you decide the floor is fine.

Why a phone alone cannot measure flatness.

Worth being blunt about this, because it is the limit of the tool. A phone measures the angle of the few centimetres of surface directly beneath it. That makes it a good instrument for slope and a useless one for flatness on its own.

Set a phone down in a shallow dish-shaped hollow and it will read 0.0°, because the bottom of a hollow is horizontal. Set it on the crest of a hump and it reads 0.0° there too. The floor is 8 mm out and the instrument is telling you it is perfect, and it is not lying — you asked the wrong question.

The straightedge is what supplies the missing length. Lay a 2 m straightedge on the floor and it bridges from high point to high point; the gap underneath is the deviation, and you measure that with a tape, a feeler gauge, or a stack of coins if that is what you have. Then rest the phone on the straightedge and it gives you the slope of that whole 2 m run rather than of one small patch.

Use the two together and you get both numbers from one setup. A long spirit level works as the straightedge; so does a length of aluminium box section or a factory edge of MDF, provided you have checked it is actually straight by flipping it end for end and seeing whether the gaps move.

Surveying the room.

1. Clear it properly. Old covering up, adhesive ridges scraped, screed nibs knocked off, floor vacuumed. A survey done over grit measures the grit.
2. Chalk a grid. Roughly one metre squares, lettered one way and numbered the other, so you can write down "C3: 5 mm hollow" and find it again an hour later. Sketch the grid on paper.
3. Walk the straightedge along every grid line. Both directions, and both diagonals in any area that looks suspect. Note the biggest gap on each run and whether it is a hollow in the middle or a rock at the ends — a straightedge that see-saws is sitting on a hump, which is a different fix from a dip.
4. Read the slope with the phone at the same time. With the phone on the straightedge, switch the units to millimetres per metre and note the reading and which way it falls. Doing both measurements at each position saves crossing the room twice.
5. Pay attention to the edges. Perimeters, doorways and the area in front of a hearth are where slabs most often dip or lift, and they are also where flooring failures start.

Twenty minutes and a sheet of paper gives you something no amount of standing in the doorway squinting will: a map with numbers on it.

Finding level across a whole room.

Slope readings taken with a 2 m straightedge tell you the local fall, but they do not directly tell you how much lower one corner of a 5 m room is than another. Errors accumulate if you try to add them up.

With a rotary or cross-line laser, this is trivial: set it up, and measure down from the beam to the floor at each grid intersection. The differences are your heights, all referenced to one plane.

Without a laser, the old water level still works and costs nothing. A clear hose filled with water, held at two points in the room, will show the same water height at both ends because water finds its own level, and it does not care about corners or intervening walls. Mark a pencil line at the water surface at each position and you have a genuine level datum around the whole room. It is slow, it is fiddly, and over a long room it is more accurate than any phone.

The phone's job in this method is the sanity check. Once you have marked your datum, rest the phone on a straightedge spanning between two marks and confirm it reads zero. If it does not, one of your marks is wrong.

Reading the survey, choosing the fix.

With the map in front of you, the decision is usually clear.

Isolated dips, under about 5 mm

Patch them. A trowel-applied repair or feather-finish compound fills local hollows without committing you to flooding the whole room. Prime first, and respect the product's minimum feather thickness.

General waviness across the room

This is the case for a pourable levelling compound. It flows out under its own weight to a flat plane, so it fixes flatness and, incidentally, level. Depth is set by your high point: the compound has to cover the highest spot by at least the product's minimum thickness, often 2 to 3 mm, which is why finding the high point is the whole reason for the survey. Everything else in the room is deeper than that.

Humps

A hump has to come off, not be buried. Grinding a proud area down is cheaper than pouring enough compound to bury it, because burying a 10 mm hump means 12 mm of compound across an entire room — which is a lot of bags and a lot of weight.

A suspended timber floor that moves

Compound is not the answer on a deflecting timber floor; it will crack. Screw the existing boards down properly first, then overlay with plywood or hardboard, or pack the joists if the problem is structural. Fix movement before you fix flatness.

A big overall slope, but a flat surface

If the floor is flat and simply runs downhill, ask what actually needs level. For a floating floor, very possibly nothing, and you can lay straight onto it. For a kitchen, level the units' legs rather than the floor. Pouring a wedge of compound across a room to correct a slope is expensive, heavy, and often solving a problem nobody had.