The Babysitting Collective
Private operations platform for coordinating bookings, approved babysitters, and day-to-day workflows.
Overview
The Babysitting Collective needed a shared system for managing access and day-to-day operational workflows. I worked directly with the founders to translate changing business requirements into a private production application for administrators and babysitters.
I designed and built the platform end to end, covering product decisions, application architecture, relational data modelling, authorization, frontend and server implementation, testing, migrations, deployment, and ongoing iteration.
What I built
- Role-specific workspaces for the internal team and approved babysitters, including onboarding, profiles, protected navigation, and carefully scoped access
- Structured workflows for booking coordination, recurring care, changes, and post-booking administration
- Contextual communication and notifications that keep the right people informed at the right stage
- Reviewed timesheet, payout, and invoicing workflows with clear operational ownership
- Operational views that bring exceptions and pending work together instead of leaving them across disconnected tools
Product decisions that shaped it
Keep people in the matching process
The service is relationship-led. Turning it into a first-come-first-served marketplace would make the software simpler, but the care less thoughtful.
Decision
The platform supports considered matching and human judgement instead of trying to replace them.
Make privacy part of the workflow
Childcare operations involve information that should only be available to people with a legitimate reason to see it.
Decision
Access is staged and role-aware, so people see only the information their current responsibility requires.
Treat booking changes as commitments
Once a booking has been agreed, an important change affects real people rather than just a row in a form.
Decision
Consequential amendments are explicit, traceable, and brought back to the people who need to confirm them.
Keep consequential approvals human
Automation should reduce repetitive work without silently making judgement calls that affect people or money.
Decision
The platform prepares and validates operational work while leaving consequential approvals with the responsible person.
The engineering behind the calm
Enforce trust at more than one layer
A private operational system should remain safe even if one screen or request path behaves incorrectly.
Decision
Access rules are enforced in both the application and data layers, with automated tests checking that each user type can reach only what it should.
Make important actions complete or not happen at all
Some operational actions affect several connected records. A half-finished change would create real confusion for the team.
Decision
Critical workflows validate the current state, complete as one reliable change, and leave durable history where it matters.
Keep supporting services off the critical path
A successful operational action should not be undone because a separate delivery provider is briefly unavailable.
Decision
Supporting work is separated from the core action and designed to retry safely without duplicating what already succeeded.
Prevent conflicting work at the source
Two people working at the same time should not be able to create an impossible scheduling outcome.
Decision
Core scheduling rules are enforced centrally and tested under concurrent actions rather than left to a warning in one browser.
Treat production data as something that cannot be recreated
Once a system holds real operational history, convenience is never a good reason to risk that history.
Decision
Production changes follow a forward-only, tested release process designed around preserving live data.
Working with the client
Requirements often began as business-language descriptions rather than implementation specifications. I worked with the founders to surface edge cases, separate concepts that had become conflated, explain trade-offs, and stage changes they could review against real workflows.
Private client system
The production application contains sensitive operational data, so I do not link to it here or publish client records, internal screens, or production-derived examples. Any public material is sanitized or fictional.
