The Spreadsheet Is Running More of the World Than We Admit

Somewhere this morning, a hotel manager opened a spreadsheet to check occupancy. A football analyst added another row of match data. A small manufacturer updated an order forecast. A finance team reconciled numbers before somebody senior saw them. A school administrator worked through unpaid fees. A researcher cleaned observations collected in the field. Somewhere else, a business owner opened a file called something like FINAL_BUDGET_v7_UPDATED.xlsx, changed three cells and emailed it to somebody who immediately saved another copy. None of this sounds particularly remarkable. That may be precisely what makes the spreadsheet remarkable.

Few pieces of business technology have become so ordinary while retaining so much power. The spreadsheet sits in an unusual space between document, database, calculator, modelling environment and programming tool. It can contain a shopping list or determine assumptions behind a multimillion-pound investment. It can record football results, calculate salaries, forecast cash flow, organise a wedding, monitor hospital supplies or help decide how many products a factory should make next month. The underlying object barely changes: rows, columns, cells and formulas. What changes is the piece of the world we decide to squeeze into them.

Its ancestry stretches back beyond the computer. Businesses were organising information in rows and columns long before anyone clicked a cell. Accountants used paper ledgers because the grid itself solves an important human problem: it gives information an address. Put categories in one direction and observations in another and suddenly relationships become easier to see. The computer did not invent that logic. What software eventually did was make the grid responsive. Change one number and another could change automatically. The ledger stopped merely recording the world and began modelling it.

That transition helps explain why VisiCalc, released for the Apple II in 1979, occupies such an important place in personal-computing history. The early personal computer could easily be dismissed as something for enthusiasts until software demonstrated that a machine sitting on a desk could perform work businesses already understood. Recalculating a financial model no longer meant manually rewriting columns of figures. Alter an assumption and the consequences could ripple through the sheet. The spreadsheet did not merely find a home on the personal computer; it helped give businesses a reason to want personal computers in the first place.

Lotus 1-2-3 followed, becoming closely associated with the IBM PC era, before Microsoft Excel eventually became the name almost synonymous with the category. Yet the deeper story is not a sequence of software companies winning and losing market share. It is the extraordinary persistence of the underlying interface. Computing has travelled from beige desktop machines to laptops, smartphones, cloud infrastructure and artificial intelligence, while millions of people still understand immediately what it means to put January in one column and February in the next.

Perhaps that durability comes from the spreadsheet's refusal to decide exactly what it is for. Most software arrives with an opinion. Accounting software expects accounts. Customer relationship software expects customers. School-management software expects pupils, fees and attendance. A spreadsheet arrives largely empty. The user supplies the ontology. Column A can be customers, cattle, countries, complaints, chemical samples or championship results. The same freedom that makes the spreadsheet primitive compared with specialist software also makes it extraordinarily adaptable.

That adaptability has allowed spreadsheets to colonise the gaps between formal systems. A company might have sophisticated finance software, a customer database and an enterprise resource planning system, yet somebody will still export information from one of them into Excel because the official system cannot answer Tuesday afternoon's peculiar question quickly enough. A manager wants three scenarios by four o'clock. An analyst needs to combine two datasets that were never designed to meet. Operations wants a temporary tracker while the proper system is being built. The spreadsheet appears because organisations constantly encounter problems that their permanent technology was not expecting.

This makes spreadsheets interesting indicators of organisational life. Find a heavily used spreadsheet and you may have found a process the official organisation chart does not show. Perhaps one employee maintains a workbook every Friday that half a department quietly depends upon. Perhaps information is copied manually from one system into another because the systems do not communicate. Perhaps an enormous file containing macros written by somebody who left three years ago has become essential infrastructure. The spreadsheet can reveal not simply what a business measures, but where its formal systems end and human improvisation begins.

That improvisation can be immensely valuable. A small business does not necessarily need an expensive software platform for every process. Someone who understands the business can construct a workable stock system, sales forecast or cash-flow model in an afternoon. In parts of the world where businesses operate with limited technology budgets, the combination of a laptop and spreadsheet can provide capabilities that would once have required specialist systems. The same basic tool can travel from a market trader in Nairobi tracking stock to an analyst in New York modelling an acquisition, or from a school bursar in Kampala monitoring fee payments to a logistics business in Singapore examining delivery performance. The businesses are radically different; the grid is recognisable.

And spreadsheets travel unusually well between organisational levels. The person entering information might be relatively junior, while the number eventually produced by the workbook can reach a board meeting. Several transformations may occur along the way. Thousands of transactions become hundreds of rows, which become a pivot table, which becomes a chart, which becomes one figure on a presentation slide. By the time the number reaches the person making the decision, much of the machinery underneath it has disappeared. A beautifully simple percentage may have a surprisingly complicated ancestry.

This is where the spreadsheet's flexibility becomes dangerous. Specialist systems constrain users partly because constraints protect the system. A spreadsheet lets us type over a formula. Copy the wrong range. Sort one column without another. Put 12 where somebody else thought the unit was thousands. Hide a row. Reference the wrong cell. Forget to extend a formula when new data is added. None of these mistakes needs to look dramatic. A workbook can be wrong while remaining beautifully formatted, internally plausible and reassuringly precise.

There have been spectacular reminders of this. During the COVID-19 pandemic, Public Health England's reporting process famously lost almost 16,000 positive test results from its daily reporting after data exceeded the row capacity of an older Excel file format used in part of the process. The tests themselves had not vanished and the people had not vanished; a technical limitation inside a data-handling workflow affected how quickly their cases entered the wider contact-tracing system. A humble spreadsheet problem had escaped the spreadsheet and entered a national public-health system.

Academic research has encountered a different version of the same problem. Scientists working with gene names discovered that spreadsheet software could interpret identifiers such as SEPT2 as dates, altering data unintentionally. The problem became sufficiently persistent that the scientific community eventually changed the names of some human genes partly to prevent spreadsheet software from automatically converting them. Think about the direction of influence there: rather than software adapting perfectly to scientific language, scientific nomenclature was altered partly because of the behaviour of ubiquitous office software. A cell in a spreadsheet had managed, indirectly, to influence how parts of human biology are labelled.

Finance provides its own cautionary tales because spreadsheets are particularly seductive when modelling futures that do not yet exist. Change a growth assumption from 3% to 4%, adjust an interest rate, extend a formula and an entirely different future appears. This is enormously useful, but the mathematical neatness can disguise uncertainty. A spreadsheet cannot know whether next year's customers will behave according to the assumption typed into B17. It can only calculate faithfully from what it has been told. The result can therefore be numerically perfect and conceptually ridiculous.

That distinction matters far beyond finance. Spreadsheets encourage us to convert messy reality into categories because categories are what columns require. A hotel guest becomes an occupancy statistic. A pupil becomes an attendance percentage. An employee becomes a salary line. A shipment becomes a delivery time. A customer complaint becomes a category and status. This compression is necessary; organisations could not operate if every decision required rediscovering the full human story behind every number. But compression always leaves something behind. The spreadsheet gives us extraordinary visibility by deciding, explicitly or accidentally, what does not need to be visible.

It also creates its own form of organisational power. The person who understands the spreadsheet may understand a process better than the person whose job title suggests they own it. Knowing which assumptions drive a model, which columns can be trusted and which numbers require explanation creates a kind of informal authority. This is why the departure of one employee can occasionally produce the alarming discovery that an important business process was not really contained in the company's systems at all. It was contained partly in Sarah's head and partly in a workbook on a shared drive.

Cloud spreadsheets changed some of this without changing the essential object. Google Sheets and cloud versions of Excel made simultaneous collaboration ordinary. The spreadsheet stopped being only a file passed from person to person and became a shared space where several people could alter the same representation of reality at once. That solved the comedy of emailed attachments labelled FINAL, FINAL2 and USE_THIS_FINAL, but introduced new questions about permissions, ownership and who changed what. Collaboration became easier precisely because more people could reach the machinery.

Now artificial intelligence is beginning to alter the relationship again. A user may increasingly be able to describe the analysis they want rather than know the formula required to produce it. AI can help generate formulas, classify information, identify patterns and interrogate datasets conversationally. That potentially lowers the technical barrier between a question and an answer. Yet it does not remove the oldest spreadsheet problem: somebody still needs to understand what the numbers mean. Making analysis easier can increase the number of answers produced without necessarily increasing the number of good questions being asked.

There is a connection here to the wider machinery of organisations. In our exploration of The Hidden Operating System of a School, we looked at how attendance, payments, timetables, communication and records form connections underneath something apparently as straightforward as a school day. The spreadsheet frequently occupies exactly this hidden layer across organisations everywhere. It is where the official system meets the awkward exception, where structured data meets human judgement and where someone constructs a temporary solution that occasionally survives for fifteen years.

Perhaps this is why predictions of the spreadsheet's death have always felt slightly premature. Better databases exist. Specialist platforms exist. Enterprise systems exist. Business-intelligence tools exist. AI can increasingly manipulate data for us. Yet none possesses quite the spreadsheet's peculiar combination of visibility, flexibility, accessibility and permission to improvise. You can see the numbers, see the formulas, change the structure and build something the software's designers never anticipated. Its greatest weakness and greatest strength are almost the same thing: it trusts us.

So the spreadsheet deserves to be understood as more than office software. It is part of the connective tissue of modern economic life. It sits between workers and databases, businesses and decisions, raw events and reported numbers. It helps tiny organisations behave like larger ones and allows enormous organisations to patch holes in systems costing millions. It can reveal a pattern that changes a strategy or conceal an error that changes an outcome. It can be a temporary scratchpad, a scientific hazard, a financial model, an unofficial database or an institution's most important piece of technology that nobody thought to call a system.

Open enough spreadsheets and eventually you stop seeing rows and columns. You start seeing organisations attempting to make the world legible. Every heading represents a decision about what matters. Every row represents something that happened. Every formula describes a relationship somebody believes exists. And every blank cell leaves room for something the organisation does not know yet. That may be the spreadsheet's real achievement: not that it organised our numbers, but that for nearly half a century it has given us one of the simplest ways imaginable to turn a complicated world into something we think we can understand.