How the QR Code Became a Digital Gateway and What Founders Can Learn About IP Protection

The Afternoon Masahiro Hara Played Go — and Invented a New Kind of Patent Strategy

In 1994, Masahiro Hara was watching a colleague play Go during a lunch break at Denso Wave's Aichi Prefecture facility when the black-and-white grid on the board sparked an idea. His team had spent months frustrated by the same bottleneck: Toyota's assembly lines were running faster than conventional barcodes could keep up. A single barcode on a parts tray held a maximum of 20 alphanumeric characters. Tracking a subassembly through a modern automotive plant required scanning multiple barcodes in sequence, and a single smudge or misalignment caused a stoppage. Hara wanted one symbol that could hold an entire part number, supplier code, and destination — all scannable from any angle in under 0.1 seconds.

The Go board gave him the architecture: encode information in two dimensions, not one. Hara's team spent eighteen months solving two engineering problems that would, quietly, determine the invention's entire IP trajectory. First, they needed position-detection patterns that a scanner could find regardless of rotation — the three identical "finder squares" now in every QR code's corners, designed so their 1:1:3:1:1 module-width ratio appears nowhere else in the symbol by construction. Second, they layered in Reed-Solomon error correction, a 1960s algorithm already in the public domain, capable of restoring up to 30 percent of damaged data. The result, formalized as ISO/IEC 18004 in 1997, could carry 7,089 numeric characters — 350 times the capacity of a standard barcode.

Denso Wave filed for patent protection and received, among others, US Patent 5,726,435. What happened next is the strategic lesson this article is built around: they chose not to enforce it.

Why Non-Enforcement Was the Most Sophisticated IP Move of the 1990s

At first glance, letting a granted patent sit idle looks like a failure of IP strategy. In practice, Denso Wave executed what this article frames as The Protocol Inversion: in technologies that must achieve ubiquitous adoption to generate value, enforcing patent exclusivity breeds competing standards and fragments the market, while open licensing converts the patent from an exclusion right into a credibility marker within the dominant protocol — shifting the durable moat from the claim itself to the ecosystem that forms around the standard.

Consider the counterfactual. Had Denso Wave charged per-scan royalties in 1997, hardware manufacturers would have developed or licensed alternatives — Data Matrix, Aztec Code, MaxiCode — and the market would have fragmented across incompatible symbologies. Instead, by making the QR specification freely implementable, Denso Wave collapsed that competitive landscape within a decade. By the time smartphones arrived and QR codes became a global consumer interface — frictionlessly embedded in WeChat Pay, Google Lens, and every restaurant menu during the COVID-19 pandemic — no competitor could displace the standard. The patent expired. The moat did not.

For founders building protocol-layer technologies — authentication schemes, data interchange formats, communication handshakes — The Protocol Inversion reframes the core question: the goal is not to maximize royalty extraction from a single patent but to become the infrastructure competitors are forced to build on top of. The patent establishes priority and credibility; the ecosystem delivers the lasting return.

What the QR Code's Architecture Reveals About Patentable Layers

Hara's invention is a clean case study in layered IP because its components sit at radically different points on the patentability spectrum, and understanding the distinction is directly applicable to any founder building on top of existing algorithms.

The Reed-Solomon error-correction logic was prior art — unpatentable in 1994 because it had been in academic literature since 1960. The data-encoding tables governing how numeric, alphanumeric, binary, and kanji characters map to module patterns were similarly based on established mathematical schemes. Neither layer was novel.

The finder-pattern geometry — specifically the three-square, ratio-defined position-detection architecture — was genuinely inventive. It solved a concrete engineering problem (angle-independent scanning) with a non-obvious structural solution. That is precisely what patent law rewards: a specific, technical improvement over prior practice, not a mathematical concept in the abstract.

This distinction matters enormously after Alice Corp. v. CLS Bank International (2014), which introduced a two-step §101 eligibility test that has invalidated thousands of software and data-processing claims. Step one asks whether the claim is directed to an abstract idea. Step two asks whether the claim adds an inventive concept beyond the abstract idea itself — something "significantly more" than routine implementation.

A claim drafted as "a method for encoding data in a two-dimensional symbol to enable error-resilient scanning" would fail Alice at step one: encoding data is an abstract information-processing concept. A claim drafted as "a two-dimensional symbol comprising three finder patterns each having a 1:1:3:1:1 module-width ratio positioned at three corners, wherein said ratio does not appear elsewhere in the symbol data region, enabling angle-independent optical position detection" survives Alice because it is directed to a specific structural arrangement with a defined physical function — not to the abstract idea of storing data. The USPTO granted Denso Wave's claims on precisely that structural specificity. Founders who draft claims around the output of an algorithm rather than its mechanistic implementation are writing the kind of abstract claims that federal courts now routinely invalidate.

The Trade Secret That Anchored the Business After the Patent Expired

US Patent 5,726,435 entered the public domain when its term expired. What Denso Wave retained — and what founders often underestimate — was a body of operational knowledge that was never disclosed in the patent: the proprietary test suites, grading rubrics, and tolerance specifications used to certify print and scan quality at scale. This know-how is not reconstructable from the published specification. It was accumulated through millions of real-world scans across varied substrates, lighting conditions, and printer outputs.

The lesson is not that trade secrets are always superior to patents. It is that the two instruments should be mapped to different layers of the invention's architecture. The structural innovation that can be reverse-engineered from a physical product — like the finder-pattern geometry — belongs in a patent, because a competitor will discover it anyway and a patent at least gives you an enforcement right and a priority date. The operational knowledge that cannot be reverse-engineered — calibration data, quality-threshold curves, manufacturing process parameters — belongs as a trade secret, because disclosure in a patent application sacrifices indefinite protection for a twenty-year term.

In Waymo LLC v. Uber Technologies, Inc. (N.D. Cal. 2018), Waymo demonstrated that trade secret protection for LiDAR design files was worth pursuing even when competing patents existed, because the files embodied specific calibration geometry that no patent claim fully captured. The case settled for approximately $245 million in equity — a figure that illustrates the magnitude of value trade secrets can represent when they encode genuine operational specificity rather than merely documented procedures.

Applying the Protocol Inversion to Modern Founder Decisions

The QR code's IP arc surfaces four concrete decision points that founders building digital-gateway or interoperability technologies should work through explicitly:

1. Determine Whether Your Technology Is Protocol-Layer or Application-Layer

Protocol-layer technologies — those that must be adopted broadly to be useful at all — are candidates for The Protocol Inversion. Application-layer technologies that deliver value through proprietary implementation are not. A QR-code-based loyalty platform is application-layer; the QR specification itself was protocol-layer. The distinction determines whether open licensing accelerates or erodes your competitive position.

2. Draft Claims Against Structure and Mechanism, Not Outcome

Alice invalidations almost always strike claims that describe what a system accomplishes rather than how it accomplishes it. Before filing, every independent claim should be tested against the following question: could a court characterize this as "apply it" — applying an abstract idea using conventional steps? If yes, the claim needs a structural anchor. Concrete module geometry, specific signal threshold values, defined data-path architectures — these are the anchors that survive §101 scrutiny.

3. File Before Demonstration, Not After

The US one-year grace period creates a false sense of safety. A public demonstration — a trade show scan, a beta-user pilot, a technical blog post — starts the clock. More critically, it creates prior art in jurisdictions without any grace period, including the European Patent Office and the China National Intellectual Property Administration. Denso Wave's team filed in Japan in January 1994 before the QR code was demonstrated externally. That sequencing is not coincidental; it is the operational discipline founders should replicate.

4. Map Each Layer of the Invention to the Right Instrument

Frequently Asked Questions

Denso Wave open-licensed its QR code patents. Does that mean founders should always avoid enforcing their patents?

No — The Protocol Inversion applies specifically when your technology's value is contingent on universal adoption. If your invention delivers differentiated value to a defined customer segment without requiring competitors to implement the same standard, open licensing surrenders your exclusion right without a corresponding adoption benefit. The trigger for considering open licensing is when network effects — not product superiority — determine the winner.

How do I know whether my software invention will survive an Alice challenge?

The clearest indicator is whether your claim can be performed, in principle, by a human with pen and paper. If yes, courts will characterize it as an abstract idea implemented on a generic computer — a pattern that has failed §101 in hundreds of post-Alice decisions. The safer path is to draft claims that require specific hardware configurations, defined data structures, or measurable physical parameters that constrain the implementation. If your patent attorney cannot point to the structural element that makes the claim "significantly more" than the abstract idea, the claim needs to be redrafted before filing, not after an IPR petition arrives.

When does a provisional patent application actually help — and when does it create false security?

A provisional establishes a priority date for the specific disclosure it contains, nothing more. Founders who file a thin provisional — three paragraphs and a sketch — to "lock in the date" and then substantially broaden their invention during the twelve-month window will find that the additional claims receive the non-provisional filing date, not the provisional's date. Any public disclosure that occurred in the intervening period becomes prior art against those broader claims. File a provisional that is as detailed as you can make it at the time; the twelve months are for building the business, not for inventing new claim scope.

What operational steps constitute "reasonable measures" to protect a trade secret?

Courts assess "reasonable measures" contextually, but the baseline in most US jurisdictions — informed by the Defend Trade Secrets Act and state equivalents — includes: documented access logs, need-to-know access segmentation, signed confidentiality agreements with employees and contractors before disclosure, and written policies that identify the information as confidential. In Waymo v. Uber, the court's analysis turned partly on whether Waymo had consistently treated the LiDAR files as confidential across all touchpoints. Inconsistent labeling or unrestricted sharing — even internally — can defeat trade secret status. Implement the controls before the information is shared, not after a breach occurs.

The Enduring Lesson from a Go Board in Aichi Prefecture

Masahiro Hara's insight on that lunch break was not merely technical. He solved an engineering problem, then Denso Wave solved a strategic one: understanding that the QR code's value was only realizable if it became the standard, and that becoming the standard required giving the patent away. Three decades later, the finder squares Hara designed appear billions of times a day on screens, packaging, and boarding passes — and Denso Wave's name is on the original filing. The patent expired. The credibility did not.

For founders navigating digital-gateway technologies today, the operational translation is precise: identify which layer of your invention is structural and patent it with mechanistic specificity; identify which layer is operational and protect it as a trade secret with documented controls; and determine honestly whether your technology's value is protocol-dependent before deciding whether enforcement or open-licensing is the more sophisticated long game. Those three decisions, made early and deliberately, are the difference between an IP portfolio that defends a business and one that merely documents an invention.

This article is for informational purposes only and does not constitute legal advice. Consult qualified IP counsel for guidance specific to your invention and jurisdiction.