Spreadsheets are the most successful business software ever written, and most operations should keep using them. The question is not whether spreadsheets are good. It is whether the specific thing you are doing in one has grown past what a spreadsheet can safely hold.
These are the signals worth acting on.
1. One person is the system
There is a file only one person fully understands, and the business stops when they are on leave. That is not a spreadsheet problem, it is a continuity risk with a filename.
2. The same data is entered more than once
A new employee, customer or order gets keyed into two or three places. Every duplicate entry is a future discrepancy, and reconciling those discrepancies is now somebody's job.
3. Nobody can answer 'who changed this, and when'
When a number is wrong, the investigation is a conversation rather than a query. Once decisions with financial or legal consequences run through a file with no history, you have an audit problem waiting for an auditor.
4. Approvals happen in email
The rule exists, the approval exists, but the record of it lives in someone's inbox. This works until the person leaves or the thread cannot be found.
5. Month-end takes days
If closing the period means assembling numbers from several files by hand, you are paying for a report every month rather than building the thing that produces it.
6. The workaround has become the process
New staff are trained on the workaround. It has a name. Nobody remembers what it was working around. This is the clearest sign that the tooling stopped fitting a while ago and the organisation adapted instead.
7. Growth makes it worse, not better
Doubling the volume more than doubles the admin. Healthy systems get cheaper per unit as they scale. A process that gets disproportionately more expensive is telling you something.
What to do about it — and what not to
The instinct is to look for a system that replaces everything. That is usually the most expensive available option and the one most likely to be abandoned halfway. A better sequence:
- 01Pick the single process that costs the most in hours or errors. Not the most annoying one — the most expensive one.
- 02Check whether an off-the-shelf tool genuinely fits it. If it does, buy it. Custom software is not a virtue.
- 03If it does not fit, model that one process properly: the entities, the states, the rules, the exceptions and who is allowed to change what.
- 04Build it so it can be extended, and integrate it with what you already pay for so data is entered once.
- 05Migrate the real data, run both in parallel briefly, then retire the spreadsheet.
Then repeat with the next process. Businesses that replace their operations one process at a time tend to end up with systems people use. Businesses that attempt everything at once tend to end up with a second set of spreadsheets alongside a system nobody trusts.