The QR Code: From a Go Board and a Car Parts Factory to Everywhere

Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode!

In 1994 the Japanese auto parts company Denso Wave had a bottleneck. A standard barcode held about 20 characters, so workers sometimes scanned up to ten labels on a single box of parts. Engineer Masahiro Hara was asked for one label that could hold far more, handle kanji, and scan in a fraction of a second. Playing Go on his lunch breaks, he saw that a two-dimensional grid of black and white could carry vastly more information than a row of bars.

This episode explains what a phone actually computes when it reads a QR code. The three corner squares follow a 1 to 1 to 3 to 1 to 1 ratio of black and white chosen because it almost never appears in ordinary print. An alignment pattern lets software flatten a tilted or curved image, masking breaks up confusing clusters, and Reed-Solomon error correction rebuilds missing data. That toughness carried the code into payments, pandemic check-ins, currency, and crime.

• Hara's team fed magazines, newspapers, and packaging into a computer to find a pattern that would not occur by accident, which produced the corner marker ratio.
• Reed-Solomon math was first developed to protect data from deep space probes. At the highest setting, level H, a code still reads with up to 30 percent of its data destroyed.
• Logos in the middle of QR codes work only because the scanner treats them as damage and repairs around them.
• Singapore's SGQR replaced a counter full of payment stickers with one unified code, and monument makers in Japan, Uruguay, and Wisconsin carve codes into gravestones that link to memorials.
• Scammers have pasted fake codes over real ones on parking meters in Austin, San Antonio, and Boston, and phishing emails use QR images because security filters scan text, not pictures.