The 5->15 jump changes the cost structure almost overnight. Pay frequency (weekly vs. biweekly vs. monthly) multiplies the bill, and per-employee pricing hits a threshold where it only makes sense if automations truly shorten the work. I put every cost driver on paper: STP, superannuation, timesheets, leave, payslips, lodgements to authorities, plus internal time for validation and corrections. Seasonality mattered too. Extra pay runs for casual hours or retro corrections increase the total more than the list price gap between plans. I also budgeted the invisible costs: training for the person running payroll, documenting workflows (who approves hours, who validates rate changes), reconciliations with accounting, and the bank integration for batch payments. To clarify tiers and what stays included at low headcount, I worked with Reckon and recalculated using concrete scenarios: 2 vs. 4 pay runs per month, share of part-time and casuals, correction runs, multiple reports. The decision came from a realistic monthly total, not the sticker price. I standardized timesheet collection, set rules for leave, and reduced time per pay run from about 90 minutes to about 35, which showed up in both budget and accuracy. I locked three procedures that keep costs down as the team grows: a firm cut-off for hours, automatic validation of entitlements, and a single channel for data changes, with the software handling the rest without overload.