fancymatt
← works

· software

VTMS

Project-management software I built for my video team in Tokyo. It planned the work, then learned from it. Still in use.

VTMS, the Video Team Management System, was the project-management software I built for the video team I ran in Tokyo, at a language-learning company. It was my first real web application, and the video team still uses it.

working backwards

It started with one problem: every video had a publish date, and I wanted every step's deadline to come from that date. Each video series got a template once (shoot, edit, motion graphics, color, then language and technical checks), and every step had a default owner and a deadline counted back from publish. Each person opened VTMS to their own tasks, ranked by what was due first.

That alone made it worth building.

the work plans the work

It got interesting once we asked for a little data at the right moments. Staff started a timer when they began a task and stopped it when they finished. Over time the timers told us how long each kind of task really took. That gave us better estimates, which gave us better deadlines, which gave everyone better task lists. It also showed us which tasks took too long, and those were the ones we redesigned.

Checking worked the same way. Every issue found in review was traced to the task that caused it and sent to that person, even when one video had problems from four different people. The defect history showed which mistakes were common, and those went into the tutorials for new staff, so they got less common.

The system was built to plan the work. Before long, the work was producing the data that planned the work.

bigger than its job

The database of everything we'd made turned out to be valuable on its own. Our job used to end when a video was handed off. Then VTMS started scheduling what went out on which day, and it caught things like two videos of the same type landing in back-to-back slots. A tool built to organize the team ended up shaping what our viewers saw.

Early on I'd built an API layer, mostly because it seemed like the right thing to do. Later we wrote scripts against it that never touched the interface, and it paid for itself.

what it led to

It started in FileMaker, moved to PHP, MySQL and JavaScript, and was later ported to AngularJS. As of 2026, other people maintain it, and the bones are the ones I laid down. It's never needed a teardown. I credit the time spent on the database design, and not taking shortcuts to ship it fast.

It was also my pivot from video production to software. In 2016 I moved from running video to product management, which is still my day job. Years later I built something with the same shape again (owners, handoffs, and a system that learns from its own records), this time for a crew of AI agents.