The brief
This started with my own commute. I wanted an app that fit my routine and stopped me spending on rides when I was running late.
Most locals already know their bus stops well. The idea was to let them check on a stop without having to be there.

I spoke to 5 routine bus commuters, mostly friends and people on my route. None of them needed help finding a bus. They already knew their bus and their stop. What they wanted to know was when to leave.
Most transit apps open on a map and a search box, which suits a trip you’ve never taken. So the real question wasn’t how to show bus timings. It was what to leave out, so a trip you take every week starts with the answer instead of a search. I was building it alone, and live data had to stay inside DataMall’s call limits, so every feature also had to be worth what it cost.
Leaving out route planning
Route planning is the feature every transit app has, so leaving it out was the riskiest call. But it serves a trip you’ve never taken, and the commuters I spoke to were asking about the trips they take every week. I put saved trips first instead. Each one holds up to three boarding stops and the buses you care about, and opens by itself in the half hour before you usually leave.
Reminders come in two kinds. “Time to leave” reminders are scheduled on the phone as weekly alarms, so they fire with the app closed. “Your bus is 5 minutes away” alerts run on the server: it checks live timings every 30 seconds, follows one specific bus and sends a single push when it’s close.
The trade-off: a one-off trip still means searching, and BAAS can’t plan a route. I defined the audience by the trips they repeat, so I let the app be worse at the trips they don’t. The map shows a route; it doesn’t plan one.


Comparing the bus with a ride, not with another bus
Most bus apps stop at timings. But for a regular commuter, the costly mistake is usually booking a ride when running late, not picking the wrong bus. That was the problem I started with. So Compare puts the bus and a ride-hailing estimate for the same trip side by side, with the time difference and how much the bus saves. You can log rides you do take and see your monthly ride spend.
Savings Guard goes one step further: an opt-in pause on ride-hailing apps you pick, during hours you pick. The trade-off I didn’t see coming was trust. It needs Screen Time on iOS and Accessibility access on Android, and a bus app asking for those can look suspicious. The wording in the app has to explain exactly why it asks, and the stores need extra declarations for it.



A commuting buddy, not a utility
A neutral, grey utility look is the safe choice for transport, and it’s what most bus apps look like. But BAAS is meant to be opened twice a day, every day, and I wanted that to feel like checking in with a friend rather than a timetable. A playful brand also makes a small app easy to remember next to official ones.
The obvious cost was time: I drew the brand and the sheep mascot myself, so every illustration came out of build time. The less obvious one is that a cute brand has to work harder to look trustworthy, especially in an app that asks for sensitive permissions. So the personality stays in the brand, and the live data stays plain and readable.

Accepting timings up to 30 seconds old
The simplest build fetches fresh from DataMall on every request. It breaks down at rush hour, when many people check the same stops at once, which is exactly when the app matters most. So I designed for that moment. Each request tries a 30-second in-memory cache first, then a 45-second database cache, and only then DataMall. Busy stops are kept a little longer. When several people ask for the same stop at once, the server makes one call and shares the answer. An entry close to expiry is still served while the server refreshes it in the background.
The trade-off is freshness: a timing can be up to 30 seconds old. I accepted it because timings barely change in half a minute. For the moments it does matter, pull-to-refresh asks for data under 10 seconds old.

The mess: a test that lied, and my favourite feature
Alight alerts buzz your phone one stop before yours, even with the screen locked. They were the feature I was proudest of. My first check for them was too lenient and passed everything. When I rebuilt it to replay real bus routes, it caught 753 early alerts, and I worked them down to none.
Then preparing for the app stores showed me the real cost: it was the only feature that needed background location. That brought extra permissions, store declarations and demo videos, and each one was a reason for review to reject the whole app, not just that feature. My instinct was to polish all of it and launch everything together. But that could have meant weeks of back and forth with review while the rest of the app sat there finished.
So I cut it from the first public release and kept it on a separate branch for the next version. Before submitting, I checked the built app itself to confirm it asked for no background location. The beta now ships with none at all.
Testing also showed that on some Android phones, departure reminders silently didn’t fire unless an “Alarms & reminders” setting was on. No error, no crash, just nothing. The app now checks for it every time it opens, and shows a banner until it’s on.
All of this I’ve had to work through from Melbourne, where I now live. I can’t stand at a Singapore bus stop and check whether a reminder fires on time, so my beta testers there tell me what happens on real commutes.
What I’d do differently
- Read the store rules before building a feature, not after. Store review is a design constraint, like screen size. Every feature now starts with its permission cost: what it asks of the user and of review.
- Ship risky features on their own. Anything that could hold up the whole app goes in a later release.
- Decide early what the first version has to do. Then letting go of the rest is a plan, not a loss.
Outcome
About 10 beta testers have the app, with 30 more people on the waitlist. Next comes the public App Store and Google Play release, with alight alerts coming back in the following version.

“The minimalistic and clean UI of BAAS keeps me coming back.”
— Beta tester
