An invented accounting software migration: where the five weeks actually went

Everything below is invented: the business, the people, the dates and the amounts. It describes one migration in enough detail to show where the work actually is, which is almost never where people expect it.

The business is a small commercial cleaning contractor with fourteen staff, moving from a spreadsheet plus a separate invoicing tool to a single package. The owner expected a weekend of data entry. It took six weeks, five of which went on the old records rather than the new system, and that is the finding.

What to take away

Where they started

Sales were raised in an invoicing tool that knew nothing about costs. Costs were a spreadsheet, one row per payment, coded by a name in a column. The bank was reconciled in the sense that the owner looked at it. There was no balance sheet at all between year ends.

The trigger was not growth. It was a request from a customer for an aged statement of what they owed, which took two hours to assemble and turned out to be wrong.

The five weeks nobody planned for

Step one was supposed to be exporting the data. It became reconciling the old records, because there was nothing sound to export. The federal guide to starting a business and keeping records describes what those records are meant to consist of. The gap between that description and a spreadsheet of payments is most of this story.

Receivables. The invoicing tool listed 41 open invoices totaling 63,200. Working through them found nine that had been paid and never marked, two raised twice for the same job, and one for a customer who had disputed it eight months earlier and never been answered. Real open receivables were 48,900.

Payables. There was no payables record at all. Costs existed only when they were paid. Reconstructing what was owed at the changeover date meant collecting unpaid supplier invoices from three inboxes and a drawer. It came to 21,400, of which 6,800 was over ninety days old and two suppliers had stopped delivering.

The bank. Reconciled to the day, and the only part of the old records that was.

A loan. Repayments of 1,150 a month had been coded to a cost account for two years. The lender's statement showed the balance, and the difference between that and what the spreadsheet implied was around 24,000 of costs that were never costs.

Money held for an authority. Amounts withheld from wages had been paid over each period from the bank, and never recorded as owed in between. The balance at the changeover date had to be reconstructed from the payroll records.

That last one is why the accountant was brought in before the move rather than after. Reconstructing it correctly is not a bookkeeping exercise and the owner should not have attempted it alone. What is owed, and when, follows from the business's own obligations, which the public guidance on paying business taxes introduces.

The opening position

Once reconciled, the trial balance at the changeover date looked like this. Every figure is invented.

The owner's reaction to this document is why the case is worth writing down. Two years of running the business on a spreadsheet of payments showed no view of the 37,800 loan, the 21,400 owed to suppliers, or the 9,300 held on behalf of an authority.

The 18,600 bank balance had felt like the position. It was not. That difference is exactly what a balance sheet is for.

Setting up

The account structure was decided before anything was entered, in a single session: what the owner wanted to see on a monthly report, and what each account meant. It came to about fifty accounts. Doing this first is what turned the rest into a transfer rather than a redesign, and the reasoning is in the chart of accounts material.

Opening balances went in as one journal from the trial balance. Then the open items, individually: 39 customer invoices and 27 supplier invoices, each with its own date and reference. That took a day and a half and was the most tedious part of the project.

It was also non-negotiable. Entered as two totals, the aged reports would have shown one line each, no allocation would have worked, and the first statement sent to a customer would have been unusable.

The parallel month

Both systems ran for one full month. At the end, four differences appeared.

None of the four was a software fault. Three were things the old method could not see and one was a definition that had never been written down.

What the first proper month showed

The month's profit and loss statement came out roughly 3,000 worse than the spreadsheet version, entirely because of the three unrecorded supplier invoices and the loan split. The owner's first assumption was that the new system was wrong.

Working through it took an afternoon and changed how the business was run more than the software did. Gross margin by contract became visible for the first time, and two contracts turned out to be loss-making at the rates agreed. Neither fact had anything to do with the package. Both had been true for a year and unmeasurable.

What it cost, in effort

The first row dominates everything else. It is also the row people leave out when they estimate a migration. That work had to be done regardless: a business that cannot say what it is owed and what it owes has that problem whether or not it changes software.

The sequence that avoids discovering it late is in the accounting software guide. The monthly routine that keeps it from recurring is in the bookkeeping foundations guide.

Common questions