The Spreadsheet That Outgrew Itself

Almost every business that reaches a few dozen customers gets there on a spreadsheet, and there is nothing wrong with that. A spreadsheet is the fastest database anyone has ever built. It needs no schema, no training, no migration plan, and no budget approval. You open it, you type, and the thing you needed to remember is remembered. For a long time that is genuinely the right tool, and switching too early costs more than staying put.

The problem is that a spreadsheet does not fail loudly. It does not crash on the day it stops being adequate. It keeps opening, keeps sorting, keeps letting you type. The failure is quiet and cumulative, and it shows up in the business long before it shows up in the file.

The failure is social, not technical

The moment a spreadsheet stops working is not the moment it gets slow. It is the moment a second person needs it at the same time as the first.

Up to that point the file has one author, and that author holds the missing context in their head. They know that the blank cell in the "last contacted" column means "I called them but they did not pick up", not "nobody has called them". They know the row highlighted yellow is a customer who asked for a discount, because they were the one who turned it yellow. None of that is in the file. It is in a person.

Add a second person and every one of those private conventions becomes a guess. The second person sees a blank cell and reads it as "nobody has called them", because that is what a blank cell means to anyone who was not there. They call. The customer, reasonably, is annoyed to be asked the same question twice.

This is why the honest test is not "is the file too big" but "how many people need to agree on what it means".

The specific signs, in the order they usually arrive

Someone keeps a private copy. This is the earliest and most reliable sign, and it almost always arrives before anyone admits there is a problem. A private copy is not laziness. It is a rational response to a shared file that someone else keeps re-sorting. The moment two versions exist, neither is true.

The file has a column nobody can define. Every long-lived spreadsheet grows a column like "status" or "stage" whose permitted values were never agreed. You will find "waiting", "Waiting", "wait", "on hold", "held", and one cell that just says "?". Sorting by that column tells you nothing, which means the column has quietly stopped being data and become decoration.

History has been overwritten. A cell holds one value. When a customer moves from "interested" to "not now", the previous value is gone, and with it the fact that they were ever interested. You can no longer answer "how many people went cold this quarter", because the file only knows the present tense. This is the loss that hurts most later and is noticed least at the time.

The important information is in an email, not the file. The spreadsheet says "sent proposal". The actual proposal, what was offered, at what price, and the reply asking for a change, are in one person's inbox. The file has become an index of work rather than a record of it.

Someone has started colour-coding. Colour is the last resort of a structure that cannot express what its author needs to say. It is also invisible to sorting, filtering, counting, and every other operation that would make the information useful.

What you actually lose

It is worth being precise, because "you should get a real system" is advice, not an argument.

What you lose first is the ability to answer questions you did not think to ask in advance. A spreadsheet answers the questions its columns were designed for. It cannot tell you which customers who asked about one product later bought a different one, because nobody made a column for it and the history was overwritten anyway.

What you lose second is the ability to leave. A file that only one person can interpret is not portable, whatever its file format. The right to take your data elsewhere is treated as fundamental in data protection law, which frames it as receiving your data "in a structured, commonly used and machine-readable format" and being able to transmit it onward without hindrance [1]. A spreadsheet passes the machine-readable test and fails the structured one, because the structure was in someone's head. This is worth applying to any system you consider adopting: a platform that cannot hand you a complete, plain export of your own records is a platform you cannot leave, and any serious one will say plainly what it gives back — ours is a JSON export of every record you hold [2].

What you lose third is trust. Not dramatically. Gradually. Someone quotes a number in a meeting, someone else has a different number, and both are reading real files. After that happens twice, people stop citing the file and start citing their own memory, which is where you began.

The counter-argument, which is fair

Moving off a spreadsheet has a real cost and a real failure mode. The failure mode is replacing a file that everyone understands with a system that nobody has agreed on, and the usual cause is trying to model the business you intend to have rather than the one you have.

The way through is to move the smallest thing that is genuinely shared. Usually that is the customer record itself: who they are, how to reach them, what they have bought, and what was said. Everything else — forecasting, campaigns, reporting — can stay where it is until the record underneath is trustworthy.

It is also worth resisting the urge to capture more than you need. A system that asks for twenty fields per customer will be filled in for the first ten customers and abandoned by the fiftieth. The discipline that keeps a record usable is refusing to collect data speculatively "in case it is useful later" [3], which is a rule worth applying to your own forms long before it applies to a vendor's.

The question to ask

Not "is our spreadsheet bad". It is probably fine. Ask instead: if the person who maintains this file were unavailable for two weeks, how much of what it means would go with them?

If the answer is "most of it", the file has already outgrown itself. The tool is doing its job. It just stopped being the place the knowledge lives.