Back to articlesRoom: Discover the world
THE YORIMICHI LIBRARYBOOK 10
TechnologyText: about 4 min

How much of a QR Code can be damaged before it stops working?

The “up to about 30%” quoted for error recovery does not refer to the area of the paper. Separate the information encoded in the black-and-white pattern from the markers that help a reader find it.

Estimated time for the article text only. Allow extra time for figures, comparisons and interactive exhibits.

An original illustration inspired by the subject. Explore the article's exhibits for comparisons of quantities and mechanisms.
The main idea

QR Code correction percentages are approximate measures relative to all codewords. They do not guarantee that a certain fraction of the image area can be erased; location markers and reading conditions matter too.

SECTION 01

A damaged sign may still contain enough clues

A small scratch on a poster’s QR Code does not always prevent it from being read. But other patterns of damage may make it unreadable. The error correction described by DENSO WAVE is a mechanism for recovering data from a dirty or damaged code, not a promise that every possible kind of damage can be repaired. What is recovered is the data the code represented, not a photograph of the paper before it was torn. Go beyond the impression that “erasing a few black squares seems to be all right.” What extra information is included? What counts as one unit? Those two questions help explain its resilience.

Code observation station

Does “30%” count the same thing that a camera locates?

The amount of information recoverable, the patterns that locate a code, and the margin around it have different roles. Separate them to understand what “30%” means.

Schematic showing three roles in a conventional QR CodeThe center represents data and correction information; three corner patterns locate the code; margins surround all four sides. This is not a scannable code.
Schematic only—not a scannable codeAn original diagram with a simplified layout. No scanning experiments were performed.
  1. 1Data and correction information

    Information calculated from the original data is added to provide clues for correcting errors. This schematic explains those roles; it does not reproduce the actual arrangement of cells or the amount of information.

  2. 2Patterns that locate the code

    In a conventional QR Code, square patterns in three corners help locate the code and determine its orientation. Their role is distinct from the information used to correct errors.

  3. 3The surrounding margin

    A conventional QR Code needs a clear margin at least 4 cells wide on each of its four sides, with no text or patterns in it. This space helps distinguish the code from surrounding text and patterns. A cell is one of the small squares that make up the pattern.

8 bits make 1 codewordA unit for counting information

These 8 boxes explain the unit. They do not show 8 cells arranged in a single row inside a QR Code.

Level L
About7%
Level M
About15%
Level Q
About25%
Level H
About30%

Level M is about 15%: an approximate measure of recovery capacity relative to all codewords.

This is not the percentage of the image you may cover, or the percentage of scans that will succeed. Damage location and conditions such as printing and photography also affect the result.

Read the sources for this exhibit

SECTION 02

Add mathematical clues to the original data

QR Code error correction uses a mathematical method called Reed–Solomon coding. Correction information calculated from the original data is added so that, within certain limits, the original values can be recovered even when the received information contains errors. This is not a system of storing several complete copies of the same text. For an analogy, knowing relationships that a set of numbers must satisfy can give you more clues to missing values than receiving only some of the numbers. This illustrates the idea; it does not mean that a QR Code can be repaired using simple addition alone. The reader follows rules built into the code in advance, rather than imagining how a URL might continue.

SECTION 03

The denominator in 30% is not the area of the paper

There are four correction levels: L is approximately 7%, M approximately 15%, Q approximately 25%, and H approximately 30%. These percentages are based on all codewords, including both data and error correction information. In a QR Code, one codeword is eight bits: a group of eight pieces of information, each expressed as 0 or 1. You therefore cannot reinterpret H’s 30% as permission to cover any three-tenths of the image. The area of visible squares and the number of damaged codewords are not the same measure. Which data are damaged and the conditions under which the code is read also matter. The official explanation explicitly says that recovery may not be possible, depending on the nature of the dirt or damage.

SECTION 04

Greater resilience also needs room

More error correction information needs somewhere to go. QR Codes come in versions with different numbers of cells. The amount of data, the types of characters, and the correction level determine which version is required. Even with the same content, increasing the correction level may require a version with more cells. Now imagine keeping the printed width fixed. Fitting more cells into that width makes each cell smaller. Maximizing the single number called the correction level does not, by itself, account for print size. The official guidance also recommends choosing a level in conjunction with factors such as the environment of use and code size. Resilience involves the space needed to carry extra information too.

SECTION 05

What the three squares and the empty margin do

Look closely at a conventional QR Code and you will see large square patterns in three corners. These are position detection markers. Before correcting data, a reader has to locate the code and determine its orientation. The white margin around it also has a role. For conventional QR Codes, the official guidance specifies a clear margin at least four cells wide on each of the four sides. A cell is one of the small squares. Not every part of the pattern serves the same informational purpose. Comparing only the area of damage does not tell you its effect on the markers or on how easily the code can be distinguished from its surroundings. Two damaged patches of equal area may leave parts with different functions intact.

SECTION 06

Read the design all the way to the white border

The next time you see a QR Code on a poster or parcel, notice the squares in three corners and the empty space around it, as well as the density of its black-and-white pattern. Even an area with nothing printed on it contributes to finding and reading the code. When making a sign, this gives you a reason to check the placement requirements before filling that margin with decoration or hiding a marker behind an illustration. “30%” was a guide to correction capacity expressed in a particular unit, not a map of how much you may boldly remove. Alongside “How much is missing?” ask, “What does this part do?” and “At what size is it being viewed?” Distinguishing the ability to recover data from the clues needed to begin reading reveals a system for protecting and delivering information within a small black-and-white pattern.

Sources and verification notes

This article is an original synthesis of primary sources and explanations from specialist institutions.

Generative AI assists with research, structure and English translation. Claims are compared with sources and editorial checks are performed; independent specialist and native-English reviews have not yet been completed.

  1. DENSO WAVE | QRcode.comError correction feature ↗ (opens in a new tab)
  2. DENSO WAVEWhat is a QR Code? | Technical information ↗ (opens in a new tab)
  3. DENSO WAVE | QRcode.comQR Code standardization | Specification overview ↗ (opens in a new tab)
  4. DENSO WAVE | QRcode.comQR Code features | Resistant to dirt and damage ↗ (opens in a new tab)
  5. DENSO WAVE | QRcode.comInformation capacity and versions of QR Codes ↗ (opens in a new tab)
  6. DENSO WAVE | QRcode.comPoints to consider when determining the code area ↗ (opens in a new tab)

Sources checked:
We checked the developer’s primary sources for Reed–Solomon coding, eight bits per codeword, the approximate L/M/Q/H correction percentages, position detection patterns, and a margin of at least four cells for conventional QR Codes. We have not converted those percentages into guarantees about image area. We have not generated and scanned our own QR Codes, and we have no measurements linking illustrated damage to success or failure rates.

Rights, status, and revision history
Status
Free public edition. Preparation and revision history is retained below.
Print
0.1.3
Rights to the text and diagrams
The original Japanese text is an independently written summary. The SVG illustrations and conceptual diagrams were newly created. No external images, original source text, or audio recordings are reproduced.
How sources are used
Sources are referenced by links only. All rights to each source remain with its respective rights holders.
  1. · v0.1.2 — Checked primary sources and explanations from specialist institutions, and prepared a private preview article with a guiding question, comparisons, conditions, and references. Added an original explanatory exhibit. Publication has not been approved.
  2. · v0.1.3 — Added links explaining how to explore the new questions in v15. The existing article text, comparison conditions, sources, and source verification dates are unchanged.

For another read, another day.

Email a correction or observation about this article

PASS IT ON

Pass this discovery on.

Send it to someone curious, or save a link for another day.

Show the link to copy

Select the URL to copy it.

THE NEXT BOOK

A new question, on the next shelf.

These paths are editorial comparisons and associations. They do not claim that different subjects work by the same principle.

Choose your next article

The articles below follow an editorial selection. For connections between their subjects, see the suggested next stops above.