Under Cover

“Sorry I’m late! I got lost trying to figure out how to get from the MRT to the movie theatre.”

“Oh nooo it’s pouring - guess we’re stuck at Ion.”

“It’s so hot ah. Let’s just stay here in the aircon. I’m too lazy to explore more.”

With Singapore’s temperature rising and sudden rain storms a part of everyday life in the city, Singapore’s many sheltered passageways can be a lifesaver for OL’s (office ladies), CBD warriors, and tourists alike. Yet many people are not aware of these passageways or find themselves getting lost in the twists and turns of the underground. Navigation tools such as Google Maps often only give street-level directions, taking people on roundabout and unsheltered routes in the hot Singapore sun.

This is especially true in Orchard Road, a shopping and lifestyle hub for locals and visitors with an underground maze of routes connecting buildings across the neighborhood. While millions visit Orchard Road every year, many are unable to fully experience all that Orchard has to offer due to the difficulty of navigating the sheltered paths.

Our goal with Under Cover is to help people better navigate the sheltered passageways in Orchard (and eventually beyond!).

To do this we needed to:

1. Build awareness of Orchard’s sheltered walkways

2. Help people use the walkways to navigate from one point to another

We decided that the best way to approach this was to build a web app that would give landmark-based directions to get from point A to point B.

1. Research - Mapping Orchard’s sheltered walkways

Our first step was to figure out what sheltered paths existed in Orchard. We scoured online, spoke with our agency stakeholders from URA and STB, sought information from the malls and developers, and went down ourselves to the ground.

We realised there were a few challenges to helping people navigate the Orchard underground. Firstly, there was no detailed single source of truth mapping the passageways. Different stakeholders had different data sets and there was often difficulty sharing across private and public sector organisations.

Adjacent malls were also often owned by different developers, who were also responsible for the signage in their respective mall. As such, there was a lack of incentive to help users easily navigate to a different mall, owned by a different developer. There was also a lack of consistency in terms of what the same corridor or connection was called.

Additionally, we had hypothesised that what people really needed, rather than just geographic maps, were landmark-based directions (e.g. “turn right at Uniqlo”), that could better help people navigate indoors. However, store data was not readily available and was also in constant flux due to store turnover.

Given these challenges, we realised that it wouldn’t be a simple matter of just pulling existing data and making it more user friendly. Rather, if we wanted to move fast and build a prototype quickly, we would have to collect a lot of the data ourselves.

2. Building a low-fidelity prototype

We worked over the first few days to put together a Figma prototype that we could use to test with users. We went down to Orchard Road and decided to prototype with one route first - 313 Somerset to Centrepoint - as there were various sheltered routes one could take between the two buildings. After collecting data on the route including landmarks and pictures, we put together a prototype consisting of:

• A home screen to select different routes

• Different versions of directions

  • A list of written directions with pictures

  • Map-based directions

  • More detailed written directions

3. Verifying our hypothesis through guerrilla testing

On Day 5 we decided to conduct guerrilla testing at Orchard Road. We wanted to understand current user behaviour as well as reactions to our prototype:

1. Is navigating Orchard Road’s sheltered pathways a real problem for people? Was there a problem to solve?

2. How do people currently navigate around Orchard?

3. How do people best consume directions? What format is most useful? Do people prefer a list of directions with pictures or a map? More or less details?

4. How important is geo-tracking for user experience?

After speaking to over 10 groups of users, we were able to validate our hypothesis:

• Yes, navigating Orchard Road was a problem for many people, especially tourists and infrequent visitors of Orchard.

• There was no one-size-fits-all preference to how people preferred to consume directions. Some preferred using a map (though all wanted geo-tracking), and some preferred a list of directions. Everyone however appreciated the landmark-based directions and the pictures of the landmarks.

4. Narrowing the project scope

We decided that given the currently manual nature of route data collection we should narrow our project scope for our hackathon MVP. We chose to focus first on the Orchard MRT cluster, which spans over ten different malls and consists of a vast underground network of passageways.

5. Collecting & organising data

We then worked concurrently on collecting route data while building a working web application. We decided on breaking down the Orchard MRT into specific building nodes and mapping the routes from adjacent nodes. We could subsequently use this data to string together more complex routes.

Collecting and organising this data proved to be a fairly time-consuming and iterative exercise, as we were figuring out as we went along what made the best landmarks, how to write directions that were succinct yet understandable, and how to structure the data in a way that would be easy to port over to our web app.

6. Building a working web app

We used T3 Stack to build a Next.js web app. We also used naive depth-first-search to compute the paths between nodes (malls). To visually display the geographic data we used MapBox.

7. User Testing

Once we had a working web app we went back to Orchard to test our prototype with actual users. We wanted to understand whether 1) the app successfully could help people navigate around Orchard Road and 2) what improvements we needed to make to facilitate navigation.

During user testing we realised that having more information, even if sometimes superfluous, was better than less information. We also realised that making sure people were properly oriented at the start was very important. If people started in the wrong place, it would be extremely difficult to find their way. Based on the finding from our user test we refined our data and product.

You can find our final HFPG project below:

https://undercover.hack2025.gov.sg/

Users can input their starting and end points or click on the location on the map. They can then select different route options based on estimated time taken and degree of exposure. They can then follow directions to their target destination, either by reading a list of written directions with landmark pictures or by following a map with pictures of the landmarks.

• Mapped 30 different routes in Orchard Road

• 100% of users tested found the product useful for navigating Orchard’s underground passageways

• Feedback from our user testers!

  • “Where can we download this now!”

  • “hey I might venture orchard today/tmr with some of my friends, ok if i use undercover with them?”

• Start small when collecting data and test first. We thought we had scoped down the product enough by focusing on the Orchard MRT cluster. However this was still a lot of data, and when initially going down to Orchard Road to transcribe the routes and take pictures we hadn’t fully thought through yet the best way was to structure and format the data. As such, we had to do a lot of revisions and iterations - going back down to Orchard multiple times, taking additional pictures, rewriting the directions, renaming files, etc.

• When in doubt, go down to the ground and get the data yourself. While theoretically we could have waited to gather all the datasets and maps from the various stakeholders to put together the routes, we realised that if we wanted to move fast and build something quickly, it would be more effective to go down to the ground and map the routes ourselves.

• Collect more data

  • Expand our prototype beyond Orchard MRT cluster (e.g. to Somerset MRT cluster, other neighbourhoods in Singapore with dense underground networks like the CBD)

• Features and improvements

  • Add feedback mechanism to crowdsource updated route data from users

  • More user testing to gauge effectiveness of UX and content

• Build user base

  • Work with STB to feature Under Cover in their Visit Singapore website

  • Reach out to SMRT/Orchard Business Association to add QR code to product at key points

  • Digital marketing