It started with a problem.
Joseph-Marie Jacquard invented a programmable loom in 1804 that used punched cards to control the pattern woven into fabric. The holes determined which threads were raised, allowing complex patterns to be reproduced reliably without a skilled pattern-setter for each repetition. Herman Hollerith, working on the 1890 U.S. Census, adapted the punched card for data processing — cards represented individual census records, and a tabulating machine read the holes to count and sort responses. The census that had taken eight years to process in 1880 was completed in one year with Hollerith's system. Hollerith founded the Tabulating Machine Company, which eventually merged to become IBM.
Historical sources below ↓Then something changed.
The punched card made data encodable and machine-readable in a portable, storable format. A stack of cards was a database that could be sorted, copied, and reprocessed. This format dominated computing and data processing from the late nineteenth century through the 1970s, when magnetic tape, disk storage, and later databases replaced it. The punched card's importance is not the technology itself but what it represented: the idea that human decisions or data could be encoded in a physical format that a machine could interpret. This translation from human meaning to machine-readable representation is the underlying principle of all programming.
What a builder can learn.
Our interpretation: the punched card represents the moment when a process became separable from the person running it. When a pattern was stored in cards rather than in a weaver's memory, the pattern could be reproduced without the original weaver. When census data was stored in cards rather than paper forms, it could be sorted mechanically without a clerk reading each form. For a business builder, the equivalent is any process currently stored in a person's head that could be encoded — in a checklist, a template, a decision tree — in a form a different person or system could execute.
A door still open.
A hypothesis worth testing: if a small business chose one frequently repeated process that currently requires a skilled person to execute from memory, and encoded it in a checklist or template detailed enough for someone new to follow, would execution quality change when the checklist was used versus when it was not?
Which processes in your business require specific people because the instructions are in their head — and what would it take to encode those instructions so others could execute them reliably?
Follow the evidence.
History is sourced below. Business lessons are editorial interpretations; future possibilities are questions to test, not predictions.
Prepared from primary-source histories. First edition: 25 Sept 2026.