← All posts

Leaving spreadsheets behind without losing your history

Automation · 23 November 2026 · 9 min read

A joinery shop that has run on one spreadsheet since 2015 will not move off it, because the history feels irreplaceable. A five-step migration: split the file into record, working and computation, export only the last year, run both for a month for new jobs only, give the archive an address, then name one owner.

The scenario

A joinery and cabinetry shop in western Sydney, eleven people on the tools and three in the office. The business has run on the same spreadsheet since about 2015. One file, Jobs 2024 (final) v6.xlsx, with a tab for quotes, a tab for jobs in progress, a tab for materials and a tab everyone knows not to touch because it has the formulas in it.

It works. That is the awkward part. The owner has looked at job management software twice and backed out both times, because the demo is clean and his spreadsheet is not, and he cannot see how to get from one to the other without losing years of history. So the spreadsheet survives, and everyone works around it. The quotes tab is four thousand rows deep and only the last two hundred are visible without scrolling. Two people have personal copies that have drifted. When someone asks about a job from three years ago, the answer depends on whose copy you trust.

This is not a story about how spreadsheets are bad. A spreadsheet is often the right first tool, and this shop got years of value from it. The question is what to do when what made it cheap early on, its total flexibility, becomes the reason nobody can answer a simple question.

Here is the method for getting off it without losing the history, and without a big-bang migration week.

Why people stay on a spreadsheet that has stopped working

Three reasons, and understanding them is most of the work.

The first is that the history feels irreplaceable. Every past quote, client and job note lives in that file. Anything new starts empty, and an empty system is worse than a messy one for the first month, because the messy one can answer a question about 2019.

The second is that the spreadsheet has quietly become the source of truth for things it was never designed for. A cell holds a material price someone updates. A hidden column holds the margin calculation. The real business logic is in formulas that three people half understand.

The third, the honest one, is that the spreadsheet has no learning curve. Anyone can open it. A job system asks people to change how they work, and in a small shop the person who resists loudest is often the one whose job the system improves. Pushing a migration onto that person without their buy-in is the fastest way to end up running both forever.

The rule that makes the migration safe

Before any step below, agree on one rule, and write it on the whiteboard in the office.

Nothing is deleted from the old file. It is archived, read-only, and can be searched for as long as the business needs it.

That single decision removes the reason the migration keeps stalling. You are not asking anyone to give anything up. You are deciding that new jobs start in the new place, old jobs stay where they are, and both remain answerable.

Keep the file, protect the sheets with a password the owner keeps, and stop editing it. The mistake is trying to import every historical row into the new system and inheriting ten years of inconsistent columns along with it. You do not need to. You need the last twelve months of live jobs to move, and everything before that to be searchable.

Step one: separate the three things in the file

Open the spreadsheet with two people and mark every column with one of three labels.

The first is record. Client name, job number, date, address, what was quoted. This is the information you need to look up later, and it is the part that must survive intact. These columns are worth migrating.

The second is working. The stage column, the notes about who is doing what, the internal reminders. This is live information that changes daily. It moves into the new system, but it does not need to be brought across from old jobs, because nobody is currently working on old jobs.

The third is computation. Margin percentages, material totals, anything with a formula behind it. This is the dangerous one, because it is where the business logic hides. Every formula on the sheet is a rule someone made years ago.

That category decides whether the migration succeeds. If the margin calculation came from a pricing policy, write the policy down in a sentence before you move anything. If it was adjusted to win a particular job years ago and never reset, you have found a live pricing bug rather than a spreadsheet problem.

Step two: export the last year, not all of it

Take the last twelve months of jobs and export them as a plain CSV. Not the whole file. A year is enough to test the new system, answer almost every live question, and stay small enough that a person can actually check it.

Count the rows before you import them, and count them again afterwards. If the numbers differ, stop and find out why. Duplicated job numbers and blank rows are normal in a hand-edited file, and cleaning them out is easier in the CSV than in the new system, which rejects them with an unhelpful error.

Then import, and check one thing properly. Pick five jobs at random from the CSV, and for each one open the matching job in the new system and compare the client, the total and the description, word for word. Five jobs, five minutes, and it catches the whole class of import problems where a column has shifted by one.

Step three: run both for a month, for new jobs only

For the first month, new jobs are entered in the new system, and the spreadsheet stays open in read-only mode for looking up history. Nobody does data entry twice. That matters, because double entry is what makes people abandon the new system at week three.

What does run twice for a while is the quoting step, and only for the first fortnight. The person quoting writes the quote in the new system, then does a quick sanity check against how they would have done it in the spreadsheet, and notes any difference. This feels slow, and it is doing the work. Almost every gap between the two shows up here: a delivery charge the spreadsheet always added, a surcharge for a particular material, a rounding rule nobody remembered.

Do not let this fortnight drift into three months. Two weeks of deliberate comparison, then move on.

Step four: give the history a home people will find

The old spreadsheet needs an address. A file on one person's desktop will be lost, and a file on a shared drive under a name nobody remembers will be searched for badly and then abandoned.

Use the new system's client record if it has one, and paste a summary of each old job in as a note. Client, date, description, total. Not the whole row, just enough to answer whether you have worked with these people before and what you charged. Twelve months of this takes one person an afternoon.

Everything older stays in the protected spreadsheet, with one line of instructions in the new system's notes on how to open it. This is a mundane detail and it is the one that stops the migration feeling like a loss. When someone asks about a 2018 job, the answer exists, and everyone knows where it is.

Step five: decide who owns the new source of truth

A spreadsheet has no owner, which is why it drifted. A new system needs one, in writing, and it is nearly always the job of one of the office staff rather than a committee.

Give one person responsibility for the shape of the data: what counts as a completed job, which fields are mandatory, what happens when a quote is revised. Give the same person a weekly ten-minute check of records with missing information. Not a report, just a list of jobs with a blank where something should be.

That weekly check is what most reliably keeps a new system alive, because the failure mode of every job system is not a crash. It is slow drift, where the important notes end up in text messages instead, and after six weeks the system is technically running and nobody trusts it.

The failure modes to watch for in the first month

Three things go wrong, predictably.

The parallel spreadsheet. Someone quietly starts a new one for a purpose the system does not serve, usually because it is missing one thing they like. Treat that as a feature request rather than a discipline problem, and look at it within a week. A parallel spreadsheet is a symptom that the new system is missing something small.

The second is the historical question arriving at the worst moment, on a busy Friday, when the person who knows where the archive lives is not in. That is why step four comes in week one, not week four.

The third is the person who keeps entering jobs in the old system because they were told rather than asked. No amount of training fixes that. The people who touch the file daily should be in the room when the three labels at step one get decided, and if they are not, the migration is the owner's project rather than the business's.

What good looks like after a month

Nobody opens the spreadsheet to check something the new system should know. The archive still exists, protected, opened twice in a month, both times for a job more than a year old. New quotes start and finish in one place, and the person doing them has stopped running the sanity check because the differences they were finding have all been resolved.

The weekly check produces an empty list most weeks, and when it does not, it is two or three rows rather than twenty. The personal copies that had drifted have been deleted, and if anyone asks why, the answer is that there is one place jobs live now.

The honest measure is a question for the office manager: where would you look to find out what we charged a client in 2019. If there is one answer and everyone gives it, the migration is done. If there are three answers and one of them is a spreadsheet, the spreadsheet is still the source of truth.

The last thing worth saying is that the new system will be wrong about something, probably within a fortnight. That is normal, and it is smaller than the alternative: another year of a file that only two people can navigate and one person can explain.

Common questions

How do I move off spreadsheets without losing years of job history?

Do not try to migrate everything. Import the last twelve months of jobs into the new system and keep the old file as a protected, read-only archive that anyone can still search, so both stay answerable and nobody feels like they are giving something up.

Should I import all my old spreadsheet data into a new job system?

Rarely. Old rows carry years of inconsistent columns and duplicate numbering, and cleaning them up costs more than the value of having them in the new system. Import a recent slice and leave the rest in the archive.

Why do businesses go back to spreadsheets after buying job software?

Usually because the people doing the data entry were told about the change rather than included in deciding it, or because the new system is missing one small thing they relied on. That gap quietly becomes a private spreadsheet nobody else knows about.

Want this done for you?

I build one small automation at a time for North Shore businesses, usually starting with whatever is costing you the most hours.

Start a project ↗