The History of the Spreadsheet

By Victor Hoang · Updated Oct 24, 2026

Somebody needed to try a number. They had to see what happened to every other number, and try again. On paper that meant an eraser and an afternoon. On a mainframe it meant a programmer and a queue.

The electronic spreadsheet as people use it today began in 1979, when Dan Bricklin and Bob Frankston released VisiCalc for the Apple II. The idea had earlier roots: paper accounting worksheets, a 1961 budget-simulation paper by Richard Mattessich, and a 1969 system for Bell Canada called LANPAR. A grid of rows and columns is thousands of years old. Making the grid recalculate itself, on a machine an individual could afford, is what changed office work.

This article follows that problem through six eras. For the wider story, see The History of Computers and the computer timeline. The hardware that made a personal spreadsheet possible is in the history of the personal computer and the history of the microprocessor.

A quick note on terms. A spreadsheet is a grid of cells where some cells hold values and others hold formulas that refer to those values. Recalculation means that when one value changes, every formula that depends on it updates automatically. Paper accounting grids had the grid. Only software had the recalculation.

Timeline at a Glance

The Ledger Years (c. 3100 BCE-1950s)

Scribes and bookkeepers lay out the grid, and nothing recalculates

The Metropolitan Museum of Art holds a Sumerian administrative account of the distribution of barley and emmer, dated to about 3100-2900 BCE and probably from Uruk. It also holds an Ur III "balanced account" from Drehem, ancient Puzrish-Dagan, dated to about 2039 BCE. These are precursors, and it's worth being plain about that. Scribes kept structured accounts, and one of these is titled a balanced account. They had no cells, no formulas and no idea of recalculation, and nobody called what they made a spreadsheet.

The name came much later, from bookkeeping. Richard Mattessich, in his own account of the term's history, notes that the phrase "spread sheet" appears in the first edition of Eric Kohler's Dictionary for Accountants in 1952, where it refers to a worksheet giving a two-way analysis of accounting data. Dan Bricklin says much the same about the electronic version: he and Frankston did not invent the word, and "electronic spreadsheet" stuck precisely because the program was like the paper spreadsheets he had used as a student at Harvard Business School.

So by the 1950s the grid was familiar, the vocabulary existed, and every change still meant erasing and re-adding by hand. The next era is about the first attempts to make a machine do the re-adding.

For teams today: the spreadsheet is a very old way of organizing a decision. If a process already lives on a paper grid, that grid is the specification for the first software version.

The Mainframe Models (1961-1977)

Budgets move onto computers, but only programmers can change them

In 1961 an accounting professor wrote down a proposal. Richard Mattessich, then at the University of California at Berkeley, published "Budgeting Models and System Simulation" in The Accounting Review in July 1961. Per his own history of the term, he followed it with two 1964 books, one of which, Simulation of the Firm Through a Budget Computer Program, included the program itself, written in FORTRAN IV by two of his research assistants, Tom Schneider and Paul Zitlau. Scholars quoted in that history say it anticipated the spreadsheet's use of matrices and of formulas behind each cell. It ran on a mainframe, in batch, and a non-programmer could not casually edit it.

A different problem produced a different tool. Rene Pardo and Remy Landau created LANPAR, short for LANguage for Programming Arrays at Random, in 1969, according to Daniel Power's history of spreadsheets, which says it was used for budgeting at Bell Canada, AT&T, the Bell operating companies and General Motors. Their patent application, US 4,398,249, was filed on 12 August 1970 and carries a publication date of 9 August 1983, thirteen years later. The patent is titled "Process and apparatus for converting a source program into an object program."

Bricklin is generous about this period. He says there were many row and column programs before VisiCalc, that Frankston had worked since the 1960s at Interactive Data Corporation, a major timesharing utility, and that most earlier financial forecasting tools ran on timesharing systems costing more than $1,000 a month or used card input. The pieces existed. What was missing was a version a single person could sit down at.

For teams today: a good tool that only specialists can touch stays a specialist tool. The mainframe models solved the math and left the access problem for the next era.

The VisiCalc Years (1978-1982)

A business student's daydream becomes the Apple II's reason to exist

In the spring of 1978, in Room 108 of Aldrich Hall at Harvard Business School, Dan Bricklin was daydreaming through his MBA classes. He wrote that he imagined a calculator with "a ball in its back, like a mouse" and a screen of numbers floating in front of him. That summer, on a bicycle on Martha's Vineyard, he decided to turn it into a product. In early fall he prototyped it in BASIC on a borrowed Apple II, and his work journal dates that first prototype to October 8, 1978. The mouse he had in mind gave way to the Apple II's game paddle, and then to two arrow keys and the space bar. For the mouse story, see the history of the mouse and the GUI.

Bricklin and his friend Bob Frankston, who did the programming, made a deal with Dan Fylstra of Personal Software to publish the result. They incorporated Software Arts on 2 January 1979. Frankston wrote the program in assembler for the 6502 microprocessor, and, as the Early Days page explains, VisiCalc used a variation of decimal arithmetic so money values could be represented exactly. The 6502 and its siblings are part of the story told in the history of the microprocessor.

The public debut was quiet. VisiCalc was announced at the National Computer Conference in New York in June 1979, where Bricklin recalls about 20 friends and family and two attendees they didn't know. The first real release, version 1.37, shipped in mid-October 1979, and the original Apple II version ran on a 32K Apple II. Then it took off. The Computer History Museum calls VisiCalc the "killer app" that spurred Apple II sales, noting that many customers bought an Apple specifically to run it, and that it went unpatented because software patents generally weren't issued until after a 1981 Supreme Court ruling.

What made VisiCalc special, in Bricklin's own list, was that it was interactive in a WYSIWYG way, recalculated instantly, and ran on an affordable personal machine. Frankston adds that it was not an accounting program at all, just a tool that made it possible for people to do accounting. In plain terms, a manager could ask "what if the price is 10% lower?" and watch the whole plan change in a second. For the machine under the program, see the history of the personal computer.

For teams today: the winning feature was not the grid but the loop. Change one number, see every consequence at once. When you evaluate a tool, ask how short the loop is.

The Lotus Years (1982-1987)

1-2-3 turns the IBM PC into a business machine

Mitch Kapor had written two business programs for VisiCorp before he founded Lotus in April 1982. According to a company history, he joined forces with programmer Jonathan Sachs, who spent ten months writing Lotus 1-2-3 in assembly language for the IBM PC. They aimed to make it a fast recalculator and used the 256K of memory that new PCs offered. The same history says Lotus spent more than $1 million on advertising over three months, that 1-2-3 took more than $1 million in orders in the first days after its official release in November 1982, and that nearly 110,000 copies sold at $495 each in its first nine months. Lotus finished 1983 with $53 million in sales.

Bricklin recalls that Lotus shipped 1-2-3 in early 1983, while the company history dates the official release to November 1982. Bricklin says it quickly replaced VisiCalc and every other spreadsheet of the day as the dominant spreadsheet on the IBM PC. The Computer History Museum calls it the first "killer application" for the IBM PC, a software suite built on a word processor, spreadsheet and database.

VisiCalc's makers ended up in a legal fight with their own publisher. VisiCorp sued Software Arts in September 1983, and the two settled in the summer of 1984. Then, in the spring of 1985, Software Arts' assets were sold to Lotus, which decided not to continue publishing VisiCalc. The company that wrote the first spreadsheet was bought by the company that replaced it.

For teams today: being first doesn't protect you from a better-timed competitor. Lotus won on speed, memory and the right hardware at the right moment, not on a new idea.

The Excel Age (1985-2005)

Windows hands the grid to Microsoft; this era overlaps the Lotus Years

Microsoft's own history timeline records that on 30 September 1985 the company announced shipment of Excel for the Macintosh to retail stores, a spreadsheet combining business graphics with an on-sheet database. It was a graphical spreadsheet, built for a machine with a mouse. Two years later, the 1987 timeline shows the company announcing Excel for Windows on 6 October 1987 and shipping it on 19 November.

The real shift came with Windows itself. Bricklin writes that 1-2-3 stayed the dominant spreadsheet for years, until Microsoft Excel overtook it after the emergence of Windows. The Computer History Museum's software timeline records that IBM bought Lotus in 1995, and notes that Microsoft lined up Word and Excel versions to run under Windows 3.0. For the operating system side of this, see the history of operating systems.

Excel kept the conventions that earlier tools had set. Bricklin points out that Excel 97 still uses the "/" key to enter command mode, much as VisiCalc did, because Excel had to follow the conventions of 1-2-3. A grid that began on a 32K Apple II was now on almost every business desk, and a lot of important decisions lived inside files that one person had built and nobody else had checked.

For teams today: a spreadsheet that becomes the system of record inherits obligations that a scratch calculation never had. Treat it that way before the stakes force you to.

The Shared Sheet (2006-present)

The grid moves online, and then it meets the auditors and the AI

On 6 June 2006 Jonathan Rochelle, a product manager at Google, posted that the former 2Web Technologies team had joined Google's New York office the previous year. The post said spreadsheets had real power but real drawbacks for collaborating and sharing, and announced Google Spreadsheets as a limited test on Google Labs. Its target was the "multiple out-of-sync versions that are email attachments." That sentence is the whole case for the cloud spreadsheet: one live copy, many editors. For the infrastructure behind it, see the history of cloud computing.

As spreadsheets spread, so did their mistakes. Three documented cases show the range.

• Economics, 2013. Thomas Herndon, Michael Ash and Robert Pollin replicated Reinhart and Rogoff's widely cited debt-and-growth results. Their 15 April 2013 paper found that a coding error in the working spreadsheet entirely excluded five countries, Australia, Austria, Belgium, Canada and Denmark. That error, compounded with others, accounted for a 0.3 percentage-point error in the published average for the highest debt category. Their corrected figure for countries above 90 percent debt was 2.2 percent growth, against the published figure of negative 0.1 percent. The spreadsheet error was one of three problems the paper named, along with selective data exclusion and unconventional weighting.
• Banking, 2012-2013. In a hearing on the JPMorgan Chase "London Whale" trades, the Senate record dated 15 March 2013 includes a January 2012 internal draft review of the bank's value-at-risk method. It said the computation was done off spreadsheets using a manual process, and called that "error prone, and not easily scalable." Senators pressed bank officials about that manual process at the hearing.
• Genetics, 2016. A Genome Biology study found that Excel, with default settings, converts gene names such as SEPT2 into dates. A scan of supplementary files from 18 journals published between 2005 and 2015 found that approximately one-fifth of papers with supplementary Excel gene lists contained erroneous conversions.

The newest turn is that the formula itself is optional. On 16 March 2023 Microsoft announced Copilot for Microsoft 365, saying that in Excel you can "analyze trends and create professional-looking data visualizations in seconds". That moves the spreadsheet from a place where you write a formula to a place where you ask a question. It also moves the checking problem. A formula can be audited cell by cell. A generated answer needs a different habit: ask what it was computed from.

For teams today: shared sheets fix the version problem and expose the review problem. Whoever owns a sheet that feeds a decision should also own a second pair of eyes on it.

Key People

Key Facts

• Software Arts was incorporated on 2 January 1979, and VisiCalc shipped in mid-October 1979.
• LANPAR's patent was filed on 12 August 1970 and carries a publication date of 9 August 1983.
• Lotus sold nearly 110,000 copies of 1-2-3 in its first nine months and had $53 million in 1983 sales.
• Excel for the Macintosh shipped to retail on 30 September 1985, and Excel for Windows on 19 November 1987.
• Google Spreadsheets opened as a limited test on 6 June 2006.
• About one-fifth of papers with supplementary Excel gene lists contained gene name conversion errors.

The Pattern Behind the History

The same four moves repeat from clay tablets to cloud sheets. We call the sequence the Grid to Loop Cycle. It's our synthesis of the history, not a research finding.

1. The grid comes first. Scribes, bookkeepers and accountants used tables long before software, so the layout was never the invention.
2. Specialists get the loop first. Mattessich's FORTRAN program and LANPAR recalculated, but only programmers and large organizations could use them.
3. Access decides adoption. VisiCalc won because one person could use it on a personal computer, and Lotus won again by fitting the next machine.
4. Scale creates the audit problem. Once many people share a sheet that feeds real decisions, review becomes the weak point, and each new tool has to answer it.

What It Teaches Teams Today

• Start from the paper grid. The best specification for a first version is the table people already keep.
• Shorten the loop. The question is how fast a changed number shows its consequences, not how many features the tool has.
• Separate the scratchpad from the record. When a sheet feeds a decision, give it an owner, a second reviewer and a version history.
• Check inputs and formulas. The documented failures involved a range that left out five countries, a manual data process, and default settings that changed names into dates.
• Ask what a generated answer came from. An AI answer in a cell is only as reviewable as its sources.
• Know which "first" you mean. First grid, first computerized budget, first commercial spreadsheet and first personal-computer spreadsheet are four different claims.

The spreadsheet shows how an old idea waits for the right machine, and how a tool built for one person's what-if becomes a system many people trust. Before the next big number leaves your sheet, ask who has checked the one underneath it.