What are the signs it has outgrown the spreadsheet?
Most of the internal tools I build replace something that started as a perfectly sensible file. These are the signs.
- Two people edit the same file, and one of them overwrites the other.
- There is a copy on someone's phone that is more up to date than the shared one.
- A formula broke and nobody noticed for a week.
- You cannot tell who changed a number, or when.
- Adding a column means rebuilding three tabs and the summary.
- The person who understands it is the person you cannot afford to lose.
- A manager asks a question and the answer takes someone an hour to build.
What does a dashboard give you that a file does not?
Three things: one source of truth, roles, and a screen that loads on a phone in a warehouse or a cab.
| What you need | Spreadsheet | A built dashboard |
|---|---|---|
| One number everyone agrees on | Only if one person owns the file | Yes, from one source |
| Two people editing at once | Conflicts and overwrites | Handled properly |
| Who changed what, and when | Rarely, and never reliably | Every change is recorded |
| Different views for different roles | Tabs, and trust | Real access control |
| Readable on a phone on site | Poorly | Designed for it |
| Feeds itself from your shop, bank or jobs | By hand, or by a fragile link | By a connection that can be monitored |
| Cost to start | Nothing | A build, then very little to run |
When the spreadsheet is the right answer
If one person maintains it, the numbers still match reality and nobody is waiting on it, leave it alone. Software should remove work, and a dashboard nobody opens is just a new thing to maintain.
The same goes for the tool you already pay for. If your shop platform or accounts package can produce the report, I would rather connect the two than build a screen that duplicates it.
- One person owns the file, and it is part of their job.
- The numbers still match the bank and the warehouse.
- Nobody waits on it to answer a question.
- A tool you already pay for can produce the same report.
How do I start, so the cost stays low?
I write down the six questions the owner asks every week, and build the screen that answers those. Nothing else. Small jobs start at $50, so a single report is a cheap way to find out whether this is worth doing.
After that first screen, the next ones are usually quick, because the data is already in one place and already trusted.
What to remember
- Replace the file when it stops being one person's job.
- Start with the questions the owner asks every week.
- Every figure on the screen should trace back to the rows behind it.
- If an existing tool can produce the report, connect it instead.
Sources
Primary documentation, for anyone who wants to check the details.