As a Product Engineer at Dozr, I built and shipped Instant Booking. Took the rental flow from sales-assisted to fully self-serve, and cut transaction time in half with 86% faster page loads.
Dozr is a construction equipment rental marketplace. Contractors across North America use it to find and book heavy equipment (excavators, skid steers, boom lifts) from a network of suppliers, without having to call around for rates.
I was a Product Engineer there, which basically meant I lived at the end of the software development lifecycle. Specs came in, I built them, and I made the real-time calls on what actually shipped.
Renting heavy equipment isn't like booking a hotel. You're dealing with delivery windows, insurance requirements, fuel surcharges, site conditions. Lots of ways for things to go sideways if the details aren't locked in. So it made sense that customers went through sales to book. Sales would get on the phone, confirm everything with the supplier, and close the deal. It was reliable.
The problem was that it didn't scale. As Dozr grew, sales was spending most of their time on completely routine bookings that didn't actually need a person involved. Contractors were waiting hours, sometimes a full day, just to get a confirmation on something that should've taken two minutes.
"Sales was the safety net that became the ceiling. The goal was to move routine bookings to self-serve without breaking the trust customers had built up."
And the platform itself was slow on top of that. Listing pages, checkout, all of it was sluggish. So customers were waiting on the software and the sales team at the same time.
I was heads-down in implementation, but this touched a lot of people:
Being at the end of the SDLC means you're the last person before it hits real users. That kept me pretty honest about what was actually shippable vs what just sounded good in a doc.
After launch, suppliers who opted in saw higher booking rates, which got more of them to enable it for their inventory. It kind of snowballed from there.
Specs look clean until you're in the code. You hit edge cases and API limitations that nobody planned for. A big part of the job was making those calls fast and communicating them clearly so the team could keep moving.
Performance needs to be built for, not cleaned up later. The 86% improvement didn't happen by accident. We prioritized it upfront because slow checkout breaks trust just as much as a broken flow does.
Replacing a human in the flow means the code has to actually earn it. Sales was there for a reason. Customers trusted that process. Instant Booking only worked because the reliability was genuinely there underneath it.
Want to dig into the details or talk about the trade-offs?