The Blog

Predicting MRT crowding with Traincheck

We are excited to announce a huge new feature for, launching today! It's so big, we decided to give it its own name and site: Traincheck.

Our cool new app on desktop and mobile

We live in an increasingly data-driven world, where more and more things are being measured and more decisions are being backed by information. From finance to weather, from traffic to video games, companies and startups are upending fields by using big data, statistics, and machine learning.

Waze has done an amazing job putting these methods to good effect: they tell you the best time to leave for an event, and are able to predict what traffic will look like throughout the day. What if we could do this for public transit in the Philippines, with its informal nature and lack of timetables?

Traincheck tells you the best time to take the MRT. It predicts when crowds are heaviest, so you can commute with less stress. How? A predictive model needs data to function, and historical data isn’t readily available. We took an innovative approach to collect this data.

The DOTR MRT3 CCTV site. That's a lot of four-letter-acronyms.

The DOTr has a public site where consumers can see real-time camera streams from every station of the MRT3. We wrote software that captured images from every stream at regular intervals, and we noted the time these images were captured. This gives us a “snapshot” of every station throughout the day.

Our Internal Ranking Interface

We then built the internal tool you see above. This interface lets us view images from all stations across time, and assign a congestion estimate to each image. We have a subjective scale with five levels of congestion, and collect data from multiple users.

Over time, our software can begin to form an understanding of when crowds are formed, based on these image ratings. Using this, we can use historical data to derive patterns by which we can predict future congestion levels.

(The math behind this is pretty cool. Expect to see a post from Wil or Albert, our resident data wizards!)

Now, we just hook it up to a nice UI with easy-to-use controls, user-readable graphs, and actionable suggestions… and voilà!

Screenshot of Traincheck

Traincheck is live, and you can try it for yourself at

But wait, you may ask, doesn’t it make sense to include Traincheck as part of the main app? Of course. We’ve pushed updates to our website and mobile apps, and you get Traincheck predictions as part of the commute directions you love. on the web and on Android, now with MRT prediction data

In fact, we don’t just display this information alongside directions, we use it to improve the directions themselves. We now tell you if you should leave earlier (or later), depending on what our predictive model says about the immediate future.

We’re super proud of what we’ve built, and we’ve only started. There are a lot of even more awesome things coming up on our pipeline, and we can’t wait to release them.

Until then, ride on! 🚃

This piece is cross-posted on Medium here.