One thing that helps even by hand: build the forecast from two numbers per week. Scheduled recurring visits times your average visit value, plus one-off jobs at their quoted amount. It's rough, but it looks forward, and comparing it with actual invoices each month tells you how far off it runs.
I set this up for Jobber users as a nightly sync into your own Google Sheet, with forecast vs actual by week, month and quarter. Recurring cleaning pricing is the part I'd test on real data before promising anything. Open to trying it on your account? Read-only, and you can revoke access any time.
(Disclosure: this is my own small service.)
