Error correction levels explained: L, M, Q, H

Every generator offers four levels and almost none explain them. Picking the highest “to be safe” is a real mistake, because the two things you are trading off both affect scanning.

What the levels are

Level	Recoverable	Size cost
L (Low)	~7%	Smallest
M (Medium)	~15%	+10-15%
Q (Quartile)	~25%	+25-30%
H (High)	~30%	+40-50%

The percentage is how much of the symbol can be missing or wrong while still decoding. The mechanism is Reed–Solomon coding — the same family used on CDs and in deep-space transmission — which stores extra codewords so the original can be reconstructed from a partial reading.

Why “always use H” is wrong

Redundancy is data. More redundancy means more data, and more data means either a bigger grid or smaller modules.

Take a URL that produces a 33 x 33 symbol at level L. At level H the same URL might need 45 x 45. Print both at 2 cm and each module goes from 0.6 mm to 0.44 mm — a 27% drop in the size of the thing the camera has to resolve.

So level H buys damage tolerance and spends scanning margin. On a clean menu under good light, that trade is a loss: you made the code harder to read to protect it from damage it will never take.

Choosing by the risk you actually face

Level L — the code lives on a screen, or on protected material handled carefully. Digital invoices, presentation slides, a code inside a PDF. Smallest symbol, no tolerance for dirt.

Level M — the sensible default, and right for most printing. Menus, flyers, business cards, posters indoors. Survives ordinary handling without inflating the symbol.

Level Q — material that will be handled, folded, or lives outdoors. Loyalty cards, packaging, labels, anything in a pocket. Also worth it when print quality is uncertain.

Level H — required whenever there is a logo, since the logo is the damage. Also correct for industrial labels, equipment tags, anything exposed to weather or abrasion, and any surface that will be scanned in poor conditions.

The logo case, specifically

A centred logo covering 20% of the width removes 4% of the area. That sounds comfortably inside level M’s 15% — and it often is not, because recovery is measured in codewords rather than area, and a contiguous block of loss is harder to recover than the same amount scattered.

Raise to H whenever a logo is present. Not as a precaution: as the thing that makes the logo work at all. There is more in Adding a logo without breaking it.

The interaction with size

The two decisions are linked, and people usually make them separately:

• Raising the level makes the symbol denser at a fixed physical size. If you go from M to H, the code needs to be printed larger to keep the same module size.
• If you cannot print larger, raising the level can make scanning worse, not better.

The rule that ties them together: aim for at least 0.5 mm per module in print. Choose the level first, then check the size supports it — see the minimum print size for the numbers.

What it does not protect against

Error correction repairs missing or misread modules. It cannot help with:

• A covered finder pattern. Recovery happens after the code is located; if the three corner squares are damaged, there is nothing to recover into.
• Low contrast. The decoder cannot tell dark from light in the first place, so there is no partial reading to repair.
• Too small to resolve. Same problem: no data is being read.
• A dead destination. The code decodes perfectly to a URL that no longer exists.

All four are more common causes of failure than damage, which is why reaching for H rarely fixes a code that will not scan.

The practical answer

Use M unless you have a specific reason not to. Use H when there is a logo, or when the surface will genuinely take a beating. Use L only on screens.

If a code will not scan, the level is almost never the cause — start with the twelve usual causes instead.