Spreadsheets to Software, Step 1: Find the Process Living in Someone's Head

Most people start this conversation by counting spreadsheets. That is the wrong end of it.

The spreadsheet is not the problem. It is the visible part of a process that nobody ever wrote down, and it is usually the smallest part. The rest lives in one person’s head: which export to pull, what to fix after you pull it, who to call when a number looks off, and when it is safe to send.

You find out how much of it lives there the first week that person takes a real vacation.

The tell

Ask whoever runs the process to walk you through it. Then watch for the moment they say “and then I just.”

“And then I just check the dates against the other tab.” “And then I just email Marcus if the totals don’t match.” “And then I just know that account bills differently.”

Every “I just” is a step that exists nowhere except in that person’s judgment. Those steps are the reason the process cannot be handed to someone else, and they are the reason a new hire takes three months to get it right. They are also the part software has to account for, which is why the counting-spreadsheets version of this exercise misses so much.

When we worked with O&S Construction, the spreadsheets were the obvious symptom. The real work was in the steps between them, the ones that only existed because someone had learned them the hard way.

Four questions worth asking first

Who is the only person who knows how this actually works?

If there is exactly one name, you have found the thing worth fixing. Not because that person is a problem, but because the business is carrying a risk it has not priced. If there are three names and they describe the process differently, that is the same finding in a different shape.

What happens between the export and the report?

This is where the undocumented work hides. The export is a file. The report is a file. The interesting part is the forty minutes in between, and almost nobody can describe it accurately from memory. Have someone narrate it while they do it.

Where does the process leave the spreadsheet?

Almost every one of these processes runs through email, a phone call, or someone walking over to a desk to ask. Those steps do not show up when you look at the file, and they are frequently the slowest part. A process that takes four minutes of typing and two days of waiting for an approval has an approval problem, not a spreadsheet problem.

What does it cost when it goes wrong?

Not the worst case. The normal case. An invoice that goes out with the wrong number, caught a week later, costs a phone call and some goodwill. The same error caught at year end costs considerably more. If nobody can answer this, that is worth knowing too, because it means the errors are not being tracked.

What you do with the answers

Write the process down. All of it, including the “I just” steps. That document is worth something on its own, before anybody builds anything, because it is the thing that lets a second person run the process.

It also tells you the honest answer to whether you need software. Sometimes the write-up shows a process that is fine and just undocumented. Sometimes it shows four hours a week of copying that could be one automated step. Either way you now know, instead of guessing.

If it turns out to be the second one, that is roughly where we start too. We do a discovery pass with the people who actually do the work, write down what they know, and pick the one workflow worth automating first. You can read more about how that works, or get in touch and we can talk it through.