Poo Patrol Ops

The problem

The marketing site solved the front door. Everything behind it — who worked which yard, who still owes money, whether last month's ad spend actually produced jobs — lived across Jobber, two ad platforms, a spreadsheet, and somebody's memory. Answering a simple question meant opening four tabs and doing arithmetic.

What got built

A nineteen-model Postgres schema and a sync engine that pulls Jobber on four staggered morning crons, joined against Meta and Google ad spend, GA4, and Search Console. On top of that sits an insight and alert layer with configurable rules, recipients, and snoozes, so the app tells staff what changed instead of waiting to be asked. It shares a deployment with the public site but lives behind its own JWT-gated login.

Details worth naming

• Nightly Jobber sync split across four staggered crons to stay inside API rate limits.
• Receivables tracking — who owes what, how long it has been outstanding, and which invoices to chase.
• Productivity and payroll: timesheets, per-visit rates, and bonus rules computed from completed work.
• Re-clean tracking, a domain-specific quality loop for yards that needed a return trip.
• Configurable alert rules with recipients, snoozes, and an event log, delivered by email.
• Marketing view joining ad spend to booked revenue, so cost per job is a number rather than a guess.