Launch Day Morning
I woke up at 5:30am the morning of our first live event. Coffee, dashboard, phone in hand. Health checks were green. The previous night's deploy was holding. Cost graphs were flat. By 9am, real employers would be sitting in front of real workers, using KinTrades to match for actual jobs.
The thing I felt most acutely that morning was how much of the work I was about to rely on had happened in the four months before — and how the work for April had been making sure I could see what was happening once people started using the platform at scale.
Shipping to production is a milestone, not a finish line. April was about operating.
Building Is the Easy Part
April shifted the work from building to operating — monitoring health, tracking adoption, responding to issues, and keeping the platform reliable as real usage patterns emerged.
Operational Dashboard
We built an internal monitoring dashboard that provides a single-pane view of platform health. It's restricted to the operations team and covers:
- Service health — status of all platform components, checked on a recurring schedule
- Health history — stored over time so trends are visible, not just current status
- Adoption metrics — registration and activity tracking for both workers and employers
- Recent activity — a real-time feed of platform events
- Cost tracking — cloud spending by resource, so there are no surprise bills
Alerting
The dashboard integrates with the cloud provider's alerting system to surface active alerts filtered by environment. Alert rules are defined in infrastructure-as-code templates — version-controlled, reviewable, and reproducible across environments.
Alerts cover API response times, error rates, and resource utilization. The goal is to know about problems before users report them.
Deployment Tracking
Every deployment is tracked — when it happened, which environment, what changed. Combined with health history, this makes it possible to correlate a deployment with a change in error rates or performance. When something goes wrong, the first question is always "what changed?" Deployment tracking answers it immediately.
End-to-End Testing
April added a comprehensive E2E test suite covering critical paths: registration, profile creation, job posting, application, and the hiring flow. These tests run against the pre-production environment and validate real database state.
The E2E suite catches issues that unit and integration tests can't — environment-specific bugs, configuration mismatches, and interaction patterns that only emerge in a fully deployed system.
Community UAT: $50 or Two Hours of Mentorship
Automated tests catch a lot. They don't catch what happens when a real first-time user opens the platform and tries to do a thing they've never done before. We needed humans — more humans than my husband Ed, our intern, and me. Ed had been showing up for testing nights for months and was honestly tired of it (he hates computers; this whole thing was his personal nightmare expressed as a labor of love). Our intern was excellent but was one person.
So in April, with the hiring event a few weeks out, we opened community user-acceptance testing. We posted an offer to students in the community: show up for a testing session and you can take $50, or two hours of mentorship. They had to choose one.
Every single student who showed up chose the mentorship.
That blew me away the first time it happened and kept blowing me away every session after. These were young adults exploring their futures. The $50 was real money. They picked the mentorship anyway — every time.
I taught them what a UAT is. What the difference is between a unit test and an acceptance test. Why companies pay people to do this kind of work. What good feedback looks like, and how to write it down so a developer can act on it. They got that as their compensation. We got real testing — real first-time-user testing — from people who didn't know our product, didn't know our assumptions, and had no reason to be polite about the parts that didn't work.
Their feedback found things our internal testing couldn't have. Confusing labels. Onboarding steps that felt longer than they were. Buttons that did things they didn't expect. Every issue went on the project board. Most of them got fixed before launch.
That exchange — testing for learning — was the most KinTech thing we did in the entire build. It's the model I want to keep using.
Engagement Systems
Adoption data revealed specific drop-off points where users started a process but didn't finish. We added automated reminders for incomplete onboarding and nudges for employers who registered but hadn't posted their first job. Each reminder links directly to where the user left off.
These aren't marketing emails — they're operational messages designed to help users complete actions they've already started. The tone is helpful, not promotional.
What I'd Do Differently
I'd build the cost dashboard the same week I provisioned the first paid resource, not after the platform went live. Cost visibility is one of those things that's easy to put off and very expensive to add back in. Two hours of dashboard work in December would have saved a few "wait, what's that line item?" moments later.
What Operations Teaches You
Building a platform tells you what you think users need. Operating it tells you what they actually do. Every metric, alert, and user behavior pattern informs what to build next.
The platform launched at a regional Second-Chance hiring event — the first time real employers and workers used KinTrades at scale in a live event. The operational systems built in April ensured we could monitor, respond, and learn from that experience in real time.
Four months. Over 1,000 commits. From infrastructure to a live, monitored, production platform serving skilled trade workers and the employers who need them.
The next post in this track is the one people ask me about most: how a team of three — building nightly around a full-time tech job and a family from late October through April, then transitioning to a tighter ~22-hour-a-week cadence — actually shipped and continues to operate this platform.
