From Excel to software: six signals the spreadsheet no longer holds

Almost every Italian company under a hundred employees runs at least one important process inside a spreadsheet. Not out of backwardness: because Excel is the fastest tool in the world for building something that works immediately, and for years it genuinely does.

The problem is never the spreadsheet itself. It is that the file is born for something small, grows with the company, and at some point crosses a threshold beyond which it keeps working well for whoever wrote it and badly for everyone else. That threshold does not announce itself. It shows up as six signals, and most companies already have three of them before the question gets asked.

This article lists the six signals, says how many are needed to decide, and sets out the three routes available once the answer is that the spreadsheet has run its course.

TL;DR

The six signals that a spreadsheet has crossed the threshold:

1. It has a single owner, and when that person is away the process stops.
2. Several versions circulate of the same file and it takes a phone call to establish which one counts.
3. Maintenance time has overtaken decision time: more of the day goes into keeping the file tidy than into using it.
4. It contains macros or formulas nobody can read any more, often written by someone who has left the company.
5. It contains personal data and there is no way to know who opened, copied or emailed it.
6. It has become the source of truth for a figure that also lives in another system, and the two disagree.

Three signals out of six is the practical threshold where the cost of staying exceeds the cost of moving. Below three, fixing the spreadsheet almost always beats replacing it.

In spreadsheet-to-software work handled by Obsidian Technologies, the longest part of the job is reconstructing rules nobody ever wrote down: in mature spreadsheets the exceptions handled by hand almost always outnumber the rules encoded in formulas, and those exceptions decide whether the new software gets used at all.

The six signals

1. The file has a single owner

There is one person who knows how it is built. They know the order of the sheets, which column is updated by hand and which is calculated, that row 340 gets skipped and why.

That person is competent, and they are not the problem. The problem is that a business process has acquired a single-person dependency nobody ever decided to accept: it formed on its own. The test is simple and worth actually running: if that person takes two weeks off, does the process continue or stop?

When the answer is “it stops”, the company is carrying an operational risk it has neither documented nor priced.

2. Several versions circulate and nobody knows which counts

You spot this one from the filenames. Budget_2026_v3_FINAL_okMark.xlsx is a diagnosis, not a name.

It always starts the same way: two people need to work on it at once, there is one file, so somebody takes a copy. From that moment two truths exist, and reconciling them takes manual work nobody budgets for.

Cloud sharing has reduced the phenomenon without removing it, because the moment somebody exports a copy to work offline, or to send to an outside consultant, arrives regardless.

3. Maintenance time has overtaken decision time

The most measurable signal and the least examined.

Time it for a week: how much goes into maintaining the file (pasting data, fixing formats, repairing broken formulas, reconciling against another system) and how much goes into using it to decide something.

Once the first number exceeds the second, the tool has stopped being a tool and become a second job. On processes touching several people the total reaches person-days per month, and at that point the economic comparison against software makes itself.

4. There are macros nobody can read any more

VBA written in 2013 by a contractor who has left works until it does not. The day it breaks, because of an Office update or a field arriving in a slightly different format, the company has a stopped process and nobody able to repair it in useful time.

There is a less dramatic and more frequent aspect too: a macro nobody can read also cannot be changed. So the process stops evolving. The company changes, the spreadsheet does not, and the gap between how people work and what the file says widens until somebody starts keeping a second spreadsheet alongside.

5. It contains personal data with no traceability

A file holding names, addresses, health data, staff assessments or information about legal cases contains personal data, and the GDPR requires security measures appropriate to the risk, with access control and the ability to reconstruct who did what.

On a file sitting in a network folder those capabilities are absent. There is no log of who opened it, no control over who copied it, and a copy sent by email leaves the perimeter without a trace.

This signal has a property the other five lack: it does not degrade gradually. It sits still until an access request, a complaint or a client audit arrives, and by then the problem has already happened. In companies supplying organisations covered by NIS2 the topic also arrives contractually, through the same dynamics we described for Italian software houses.

6. It has become the source of truth for a figure that lives elsewhere too

The spreadsheet was born to extend the ERP, to handle the part the ERP did not cover. Then somebody started updating the spreadsheet first and the ERP second. Then only the spreadsheet.

When the two diverge, the company has two numbers for the same thing and no rule for deciding which wins. Meetings spent arguing about which report is right instead of about the decision are the symptom, and they are the most expensive item on this list because they consume the time of the people who cost most.

How many signals are needed to decide

One is almost never enough. Every mature spreadsheet has at least one, and replacing on a single signal means paying to solve something an organisational rule could have solved.

Three signals out of six is the practical threshold. Below three it is almost always better to work on the file: lock the formula cells, move it to the cloud with controlled access, strip out dead macros, write down on one page the rules that currently live in one person’s head.

Above three, the cost of staying grows on its own and the choice is between spending now in a planned way or spending later in an emergency. And there is an asymmetry worth stating: signal 5 counts double, because its cost is not gradual.

The three routes, and when each one is right

Stay on Excel, with discipline

This is the right route more often than software vendors admit. It holds when one or two signals are present, when the process involves fewer than three people, and when the file holds no personal data.

What it means in practice: a single file in the cloud, protected formula cells, a written list of the non-obvious rules, and a trained backup person. It costs a few days and buys years.

A standard platform

This holds when the process is common to other companies: stock, CRM, leave and attendance, document management. If somebody has already written the software for that problem, rewriting it bespoke is almost always money badly spent.

The criterion is the adaptation percentage: if the standard product covers 80% or more of the process and the remaining 20% is negotiable, adopt the product. We covered the same ground in custom software or SaaS for Italian SMEs, with the selection criteria in full.

Bespoke software

This holds when the process is the way the company does its particular thing, so that adapting it to a standard product means giving up what distinguishes it. And it holds when the spreadsheet contains rules no market product knows, typically because they come from sector-specific constraints or from arrangements with particular clients.

The right scope to start with is one process, the one where the six signals cluster most densely, rather than the whole department. Converting a spreadsheet into an application is described on the Excel to custom software page.

What moving to software does not fix

Worth writing down, because disappointment at six months almost always starts here.

It does not fix a process nobody has decided. If there are two ways of doing the same thing in the company and the spreadsheet allows both because it is a spreadsheet, software forces a choice. That choice is organisational, it has to be made first, and postponing it to development is where projects stall.

It does not fix the quality of historical data. The data entering the new system is the data that was in the file, with the same gaps. Cleaning is a separate activity.

It does not remove Excel from the company, and that would not be a sensible goal. The spreadsheet stays the right tool for one-off analysis and for prototypes. What moves is the recurring process, not the tool.

Mistakes we see

1. Replacing the most visible spreadsheet instead of the most critical. The right candidate is the one with the most signals, not the one the director complains about.

2. Rebuilding the spreadsheet as-is, inside an application. If the new software reproduces the same twenty-eight columns with the same cryptic names, the company has paid to change container. The point of moving is to remove manual steps, not to frame them.

3. Not reading the exceptions. The rules that matter in mature spreadsheets live in the hand-filled cells, not in the formulas. Skipping that reading produces software that covers the normal case and that nobody uses in the real ones.

4. Starting with the whole department. Big projects fail for calendar reasons before technical ones. One process at a time, with the first in service within weeks, keeps the consent of the people who will have to use it.

5. Switching the spreadsheet off on release day. The two run in parallel for a few weeks. The file gets retired when people have stopped opening it on their own, not when the project declares itself finished.

FAQ

How many employees justify replacing a spreadsheet?

Company size matters less than the number of people touching that file and how often. A daily process across three people justifies work sooner than a monthly process across fifteen.

Is Excel with Power Query and structured tables still Excel?

Technically yes, and it resolves a good deal of signal 3. It does not resolve signal 1 (there is still a single owner, and the required expertise actually rises), nor 5, nor 6. It is a good investment when the signals are few and the problem is maintenance time.

Access or a web application?

Access solves the single-file problem and creates a new long-term one: it is a technology with a shrinking pool of skills, and in ten years you are back at signal 4. If you are replacing, replace with something that will still be maintainable in 2036.

How long does converting a spreadsheet into an application take?

For a single, well-bounded process, think in weeks rather than months. The variable that moves the timeline most is how documented the rules are: a file whose exceptions live only in one person’s head needs a reconstruction phase, and that is the longest part of the work.

What if the spreadsheet is only used by me?

Then it stays a spreadsheet. The six signals concern shared processes. A personal analysis tool has none of the problems described here, and replacing it would make things worse.

Which process to touch first

The six signals are not there to establish whether Excel is adequate in the abstract. They are there to establish whether that file, in that process, is still the right tool now, at a company size it did not have ten years ago.

The useful thing to do does not require a project. It requires taking the three or four spreadsheets holding up important processes, counting the signals on each, and timing for one week how much goes into maintaining them. The file with the most signals and the most maintenance time is the first one, and it is rarely the one you would have picked on instinct.

If you have a spreadsheet past the threshold and want to work out whether to fix it or replace it, let’s talk. Half a day on one process produces the arithmetic for both routes, and in plenty of cases the answer is that the spreadsheet should be kept.

Further reading: custom software or a management tool for a professional firm for the professional services case, and the professional services sector page.