What If We Could Fix Barbados Traffic With Data?

Every morning during my commute, I deal with the same roundabout problem. A car is coming around, and I need to figure out: are they exiting, or staying in?

No indicator light. So I watch their trajectory, their speed, try to read their intentions. If I guess right and they exit, I can enter and save a few seconds. If I guess wrong, it's dangerous, so most times, I just wait.

I'd always figured this was just how roundabouts work in Barbados. Poor signaling, unpredictable flow, daily frustration. That's just the reality here.

Then I went to a meetup at Pelican Village on Saturday, November 22nd, 2025.

The Challenge

The turnout was smaller than usual — as persons had the option to attend via Zoom. We were there to learn about the Barbados Traffic Analysis Challenge, a machine learning competition organized by GovTech and Keleya Labs and hosted on Zindi to help the Ministry of Transport and Works (MTW).

The presentation covered the technical side: using Machine Learning (ML) to analyze video footage, extracting features from data, different classification models. A lot of it was new territory for me.

During the networking session afterward, Conrad Brits, founder of Keleya Labs, explained how the competition came together. He's been working on various projects in Barbados since moving here during COVID. He said that our highway infrastructure is great, but for some reason the speed of traffic at our roundabouts is slow when compared to other places with similar highway infrastructure. When new cameras were installed at the Norman Niles roundabout, he felt that this data can be used to predict traffic conditions with machine learning.

Then he shared his theory about what he thinks is actually causing delays.

The Signaling Problem

Conrad shared his hypothesis: people don't signal when exiting roundabouts, and he thinks that's measurably slowing down how quickly cars can enter.

Hearing him say it out loud made me realize — I've experienced this. I've seen the rare driver who does signal their exit, and I've tried it myself a few times. When someone signals, the person waiting to enter the roundabout gets those crucial milliseconds of certainty. They can go instead of hesitating.

But it's uncommon here. And without signals, every driver faces the same choice: wait to be sure, or guess and risk it.

Conrad's idea is to use machine learning to verify this with actual data. If the analysis shows that signaling improves entry rates, he can take that evidence to the MTW. Not assumptions — data. And maybe they'd update driver education to emphasize signaling at exits.

It reminded me of how jambusting became accepted practice. What started as an unofficial technique —minivans using the outside lane to go straight through roundabouts in the early 2000s — eventually became allowed at ABC Highway roundabouts as a way to ease traffic flow. Sometimes the practices that work in reality get officially recognized. Maybe signaling could be similar.

The Competition Details

The challenge itself: analyze 15 minutes of traffic video, then predict traffic conditions 5 minutes into the future. You work with four video streams from the Norman Niles roundabout, each with congestion ratings. The goal is identifying what's actually causing the delays.

There are cash prizes and exposure to international companies that recruit from these competitions. Deadline is January 26th, 2026 — two months away.

Conrad emphasized the social good that could come from this. Less time in traffic has ripple effects — people getting home to their families sooner, reduced stress, lower fuel costs, better quality of life. The improvements might seem small on an individual level, but multiply that across thousands of daily commutes and it becomes significant.

Why This Matters

There's something compelling about this challenge. The daily frustrations we accept as unchangeable might not be. Data could turn assumptions into evidence. A small group at Pelican Village on a Saturday might improve commutes for everyone on the island.

If you have any interest in data science, ML, or solving civic problems with code, check out the Barbados Traffic Analysis Challenge. It closes January 26th, 2026.

Maybe someone will prove that a simple indicator light could save thousands of Barbadians time every day.