Your business runs on a spreadsheet. Here's when that stops working.

Let's be fair to spreadsheets first. They're cheap, flexible and everyone knows them. Plenty of good businesses run on one. A consultant on Hacker News summed up a decade of client work like this: the business world runs on "insanely complex spreadsheets", built from lots of small decisions that each made sense at the time ^[1].

The problem is that small errors are almost guaranteed.

What the research says

Raymond Panko, who has studied spreadsheet errors for decades, reviewed audits of real spreadsheets used in organisations ^[2]:

• 94% of the audited spreadsheets had errors.
• On average, 5.2% of formula cells were wrong.
• Experience doesn't save you. In his lab studies, people with hundreds of hours of spreadsheet experience made errors at about the same rate as beginners.

And then there are limits nobody remembers until they hit them. In October 2020, 15,841 positive COVID-19 cases in England were left out of daily reports. Data was loaded into the old .xls Excel format, which stops at about 65,000 rows, and anything past the limit was silently dropped ^[3].

Signs you've outgrown it

• One person is "the spreadsheet person", and things stop when they're on holiday.
• The same data is typed in two places, like the sheet and the invoice, or the sheet and WhatsApp.
• There are versions: final.xlsx, final-v2.xlsx, final-USE-THIS.xlsx.
• You can't answer simple questions quickly, like "how many orders are late?", without building a new tab.
• Mistakes reach customers: a wrong price, a missed delivery, a double booking.
• Customer data is in it, with no control over who can see or copy it.

What "next" looks like

It doesn't have to be a big system. Often the right step is small:

1. Fix the input. A simple form in front of the sheet stops bad data at the door.
2. Connect what's retyped. If data moves by copy-paste, it can usually move by itself.
3. Replace the core. When the sheet is the business process, a small custom app with proper permissions and history is often cheaper than the mistakes.

Keep the spreadsheet for what it's good at: thinking, modelling, one-off analysis. Just don't make it your database.