Payroll month, without the spreadsheets.
Sable & Rowe run payroll for dozens of small businesses out of an office in Leeds. The work is exacting, deadline-bound, and until 2024 it lived in three spreadsheets.
Every client's payroll lived in a master spreadsheet, a variations spreadsheet, and a checklist spreadsheet, stitched together with email attachments from clients. It worked, mostly, because two senior people held the whole process in their heads.
The failure mode was brutal. A typo in the variations sheet didn't announce itself when it was made; it surfaced on payday, as someone's wrong pay packet and an awkward phone call. The two seniors re-checked everything, which meant payroll week swallowed them whole, and holidays needed planning around it.
Before writing any code I sat through two full payroll runs, one quiet, one horrible. The horrible one was the useful one. Watching where the team hesitated, re-checked or rang a client told me more than any requirements document, and it settled the design principle: the tool should mirror the checklist they already trusted, not replace it with my idea of a workflow.
So the tool is, deliberately, a checklist with a database behind it. Each client's run walks the same steps the team already followed. The difference is that bad data can't get in quietly: variations are validated at entry, against last month's run, and anything unusual is flagged there and then, not on payday. That flag is the feature that pays the rent.
Day one wasn't a migration cliff. The tool imported the existing spreadsheets as they were, so the team ran January half-in, half-out, and moved fully across when they trusted it, not when I said so. Every change is logged with who and when, which for an accountancy practice isn't a feature, it's the point.
The run that used to fill a day fits in a morning. The two seniors stopped re-checking every line, because the tool checks the lines, and they now review the flags instead. One of them took a two-week holiday in payroll week last summer. That sentence was impossible in 2023.
Hiring changed too. Payroll used to take months to hand over; a new starter now runs their first supervised payroll in week one, because the process is in the tool, not in someone's head.
I'd push harder, earlier, on retiring the third spreadsheet. I let it live alongside the tool for two months as a comfort blanket, and it quietly reintroduced the exact class of error we'd killed. Half a migration is a second system, not a safety net.
Payroll week is now just a week. The people who carried it in their heads got their heads back.
the audit trail has settled two client disputes since launch
next case study