In 1994 an engineer named Masahiro Hara was working at Denso Wave, the arm of the Japanese supplier Denso that built parts for Toyota. His problem was dull and very real. A standard barcode held about 20 characters, so a single crate of parts needed several labels, and workers kept stopping to scan each one in turn.
Hara wanted a label that held far more and could be read quickly from any angle. The idea arrived during a break, while he was looking at the stones on a Go board. Today the square is everywhere, from restaurant menus to boarding passes, and payment screens lean on it hard. A checkout page at a platform such as x3bet casino can show a code that a banking app reads in a second, so nobody types a long account reference by hand.
How a Go board became a data grid
Go is played on a grid of lines where black and white stones mark positions. Hara saw that a flat grid of dark and light cells could store information in two directions at once, where a barcode only uses one. Each cell, called a module, is a single bit, either dark or light. Even the first version, a tiny 21 by 21 square, holds 441 of them, and the cleverness lies in how those bits are arranged and repeated. Hara later said the Go board gave him the picture, while the maths of correcting errors came from the team around him. The grid is also built to survive abuse, and a few choices did most of the work.
- Three large squares in the corners tell the scanner where the code starts and how it is rotated.
- The squares use a stripe ratio of one to one to three to one to one, picked because it rarely appears in print.
- Reed-Solomon error correction lets a damaged code still decode.
Hara’s team reportedly scanned piles of printed material to find that ratio, so a scanner would almost never mistake a headline or a logo for a code.
| Version | Grid size | Roughly what it holds |
| 1 | 21 by 21 modules | 41 digits |
| 10 | 57 by 57 modules | 652 digits |
| 25 | 117 by 117 modules | 2,800 digits |
| 40 | 177 by 177 modules | 7,089 digits |
Why the corners matter more than the middle
Picture a delivery driver holding a phone at a slant over a crumpled parcel label in a rainy loading bay, with one hand free and about two seconds to spare. The camera does not need the whole picture to be perfect. It looks for the three corner squares first, works out the angle and the scale, and only then reads the data cells. That is why a code scans upside down or on a curved bottle.
The fourth corner has no big square on purpose. A smaller alignment pattern sits near it, which helps the software correct for warping on larger codes.
How much damage a code can take
Error correction comes in four levels, and the designer picks one when printing. Level L restores about 7 percent of lost data, M about 15, Q about 25 and H about 30. Higher levels cost space, because the extra bits are spent on repair instead of content. A printer of food packaging might pick level H for a code that will be smudged by condensation, while a poster seen from a metre away can stay at L and keep the grid coarse and easy to read.
This is the reason a logo can sit in the centre of a code and still scan. The logo simply destroys some modules, and the repair data covers the loss. Some practical limits are worth knowing.
- A scratch that hits a corner square is far worse than one across the data area.
- Low contrast, such as grey on cream, fails more often than damage does.
- Very small prints need a simpler, lower version of the code.
Why a patent holder gave the idea away
Denso Wave held the patent but chose not to enforce it, and published the specification openly. That decision is a big reason the code spread beyond car factories. Phones with cameras and cheap apps arrived years later, and nobody needed a licence to build a reader. Adoption in Japan came first, with mobile phones reading codes from magazines in the early 2000s. The rest of the world woke up much later. In 2020 menus, tickets and contactless payments pushed the grid into daily life almost overnight in Europe and North America, and many people scanned their first code at a restaurant table that year without ever thinking about who invented it.
The odd part is how little the design has changed. A code made in 1994 still scans on a phone made three decades later. It started as a fix for a warehouse annoyance, borrowed its look from a centuries old board game, and now sits quietly on billions of surfaces. Good engineering is often that modest. Hara’s own summary was that the code should simply work, whether the light was bad, the label dirty, or the person holding the phone hurried.













