Is the spreadsheet costing more than a custom tool would?
Short answer
Add up the hours your team spends re-typing and checking the same facts, then the errors that get caught late. If one year of that costs more than the tool, build it. If it doesn’t, or a product you can buy already fits, don’t.
Contents 7 sections
Nobody chooses a spreadsheet to run part of their business. Someone makes a quick sheet, it works, and two years later five people depend on it and one of them is the only person who understands the third tab.
That is not a failure. Spreadsheets are the fastest way to find out what a process is. The question is narrower: has this one become more expensive to keep than to replace? That is arithmetic, and you can do it without me.
What does the spreadsheet cost now?
Two things, and only one of them is easy to count.
The easy one is hours: the time people spend copying the same facts from an email into the sheet, from the sheet into an invoice, and checking that nobody overwrote anything on the way. The hard one is errors: the ones that get caught late, after a customer, an invoice or a report has already seen them.
Start with the hours. Four numbers, and you know all of them better than I do.
| Input | Example |
|---|---|
| People who work in the sheet | 4 |
| Hours a week each, spent re-typing or checking | 3 |
| Working weeks in a year | 46 |
| What an hour of their time costs you, all in | $44 |
| Hours a year | 552 |
| Cost a year | about $24,300 |
The example is made up, apart from the hourly cost. That is Eurostat’s average for the euro area in 2025, 38.2 euros, wages plus what an employer pays on top.11Eurostat, news release on hourly labour costs, 31 March 2026: 34.9 euros in the EU and 38.2 euros in the euro area, for businesses with 10 or more employees. It has no 2025 figure for Belgium. ec.europa.eu, checked 11 November 2026. Belgium’s 2024 figure, 48.2 euros, is from Eurostat’s release of 28 March 2025. ec.europa.eu, checked 11 November 2026. At about 0.86 euros to the dollar, that is $44. Use your own figure. Belgium is well above the average: 48.2 euros in 2024, the latest year with a Belgian number. The 46 weeks leave room for holidays.
If you have no idea how many hours it is, ask the people in the sheet to tally one ordinary week. A tally beats a guess, whichever way it goes.
A week, on average, that 500 US professionals said they spend moving data by hand between emails, PDFs and spreadsheets.
Source: Parseur, 2025
Of the same people said entering data by hand had led to costly errors.
Source: Parseur, 2025
Read those as a sanity check, not an input. It is a survey by a company that sells data-entry automation, run in July 2025.22Parseur with QuestionPro, a survey of 500 US-based professionals, July 2025. The report also puts manual data entry at “an average of $28,500 per employee annually”. parseur.com, checked 11 November 2026. If your tally says one hour a week, believe your tally.
Why do the errors matter more than the hours?
Because the hours are visible and the errors are not. A wrong number in a cell looks exactly like a right one.
Fifteen years of research studies have concluded unanimously that spreadsheet errors are both common and non-trivial.
Two public cases show the kind of error that hurts. Neither involved anyone typing badly.
Positive COVID-19 cases left out of England’s daily figures between 25 September and 2 October 2020, because some result files “exceeded the maximum file size”.
Source: Public Health England, 2020
The official statement blames the file size and does not name the software. The fix was “a rapid mitigation” that “splits large files”. Nothing warned anyone while the cases were being dropped.
The second is from genetics. Excel, with its default settings, turns some gene names into dates: SEPT2 becomes “2-Sep”. A 2016 scan of leading genomics journals found the damage in “approximately one-fifth of papers” that came with Excel gene lists.33Mark Ziemann, Yotam Eren and Assam El-Osta, “Gene name errors are widespread in the scientific literature”, Genome Biology, 2016. They screened 35,175 supplementary Excel files. PubMed Central, checked 11 November 2026. Again, nobody typed anything wrong. The tool changed the data and said nothing.
That is the real case for a tool built for one job. It can refuse a bad value at the door, keep a log of who changed what, and stop two people from editing the same record at once. A spreadsheet can be made to do some of that, and almost never is.
For your own count, list the last few mistakes that got out: a wrong invoice, a double booking, a missed renewal. Next to each, write what it cost to fix, in hours, credit notes or a customer who left. Add a year of those to the hours.
What does a custom tool cost?
An internal tool from me starts at $18k and usually takes 3–5 weeks. That covers one team and one core workflow, with accounts, roles and an audit log, and your data moved over from the sheet.
Then it costs something to run. Hosting fees go to the provider. Fixes and small changes after launch are a care plan, or your own developer, who gets the code and the handover notes. Leave this out and the sum flatters the tool. Even a report from a company that sells tool-building says it plainly: “Time saved on one task can be offset by time lost to maintenance”.44Retool, 2026 build-vs-buy report, 17 February 2026: a survey of 817 builders in late 2025, including Retool customers. retool.com, checked 11 November 2026.
So when is it worth building?
Put the two sides next to each other.
- Fig.
The twelve-month line is my rule of thumb, not a law. It is short enough that the process you are building for will probably still look the same when the tool has paid for itself.
When should you not build?
More often than people who sell custom software admit.
- When a product fits your process as it is. It costs less and you can start today.
- When you are still finding out what the process is. A custom tool freezes a guess. Keep the spreadsheet a few months longer.
- When it is one person’s two hours a week. Lock the columns, add validation, give the sheet one owner. That is an afternoon, not a project.
And when to build:
- When the process is what you sell. Nothing on the shelf does it your way, and the way you do it is the reason customers pick you.
- When the interface is the problem. The people who use it every day lose time to it.
- When per-seat fees have grown past the price of owning it. Add the fees you would stop paying to the hours in the sum.
You would not be unusual. In a late-2025 survey of 817 people who build internal software, 35% said their team had “already replaced at least one SaaS tool with a custom build”, and 49% said their tool saves six or more hours a week.4 That survey was run by Retool, which sells a tool-building platform, and it includes its own customers. It tells you what builders say, not what you will save.
How do you know the spreadsheet has become the system?
If three or more of these are true, it is worth doing the sum above properly.
- More than one person edits it every day.
- There are copies with “final” in the file name.
- One person is the only one who understands part of it.
- The same fact is typed in from an email and typed out again into an invoice or a report.
- A mistake in it has reached a customer.
- A new hire needs to be walked through it, tab by tab.
- Nobody tidies it, because nobody is sure what depends on what.
What have I built?
The internal tools I have built are private, so the internal tools page shows them as recreated screens with no client names. None of them carries an hours-saved figure yet, because I do not have one I can stand behind.
The worksheet on the internal tools page runs this sum on your own numbers, against the current price.
Notes and sources
-
Eurostat, news release on hourly labour costs, 31 March 2026: 34.9 euros in the EU and 38.2 euros in the euro area, for businesses with 10 or more employees. It has no 2025 figure for Belgium. ec.europa.eu, checked 11 November 2026. Belgium’s 2024 figure, 48.2 euros, is from Eurostat’s release of 28 March 2025. ec.europa.eu, checked 11 November 2026. ↩︎
-
Parseur with QuestionPro, a survey of 500 US-based professionals, July 2025. The report also puts manual data entry at “an average of $28,500 per employee annually”. parseur.com, checked 11 November 2026. ↩︎
-
Mark Ziemann, Yotam Eren and Assam El-Osta, “Gene name errors are widespread in the scientific literature”, Genome Biology, 2016. They screened 35,175 supplementary Excel files. PubMed Central, checked 11 November 2026. ↩︎
-
Retool, 2026 build-vs-buy report, 17 February 2026: a survey of 817 builders in late 2025, including Retool customers. retool.com, checked 11 November 2026. ↩︎ ↩︎
Written by Dion Zeneli. AI helped with research and editing. Sources are linked where they are used. Read this note as markdown.