Roster

Background

ICA Woodlands handles on average more than 327k travelers daily, taking on the critical responsibility of keeping Singapore safe and secure while connecting Singapore to the world.

To facilitate operations, officers must create daily rosters for hundreds of staff across various modes of transport. Each zone has different requirements - security screening, vehicle checks, incident response - many of which change based on conditions on the ground.

It’s essential to assign officers based on their availability, skill sets, and experience to ensure checkpoints operate at peak efficiency - especially during traffic surges or incidents that require rapid redeployment.

The opportunity

Officers spend 3-4 hours per roster daily switching between files, scrolling through lists, and cross-referencing information. Here's what that actually looks like:

To assign officers to zones, planners:

1. Open the leave file to see who's absent

2. Switch to the qualifications file (11+ sheets) to check who's trained for each zone

3. Open yesterday's roster to check who just worked where (for rotation fairness)

4. Switch back to today's roster to fill in names

5. Copy-paste between files which sometimes breaks formulas, requiring another 30 minutes to fix

Why this happens:

Information is scattered across 3-5 separate Excel files, each containing up to 18 sheets. One file tracks scheduled leaves. Another tracks last-minute MCs. A third contains officer qualifications across 11+ sheets. Then there's yesterday's roster to check rotation fairness.

The information planners need exists. It's just scattered. So instead of looking it up every time, they memorize it.

“Different files have different formats… it’s not easy lah.”

“It’s inside my head, just relying on muscle memory.”

What we learned that surprised us: officers weren't asking for fancy automation - they just wanted to stop switching between 18+ spreadsheet tabs.

"We have to refer to the same reference file as we switch between files, screenshots.”

Why it matters:

• These aren't dedicated planners, they're officers with other duties:

  • supervising hundreds of junior staff

  • coordinating training

  • handling last-minute time-off requests

• 3-4 hours on rostering = 3-4 hours not spent on people management and operations.

• During traffic surges or incidents based on operational needs, officers must be redeployed quickly, but efficient redeployment requires knowing who's qualified.

• Roster mistakes create operational stress.

Velocity

What we started with

Four teams (A, B, C, D) each with team sizes of at least hundreds handle rostering independently, each with their own Excel-based workflows that haven't changed in years. Team A had built internal macros to reduce some manual work. Teams B, C and D were still fully manual spending 3-4 hours per roster.

What officers can now do

Officers can now build rosters in under an hour instead of 3-4 hours by seeing all officer information - qualifications, availability, last assignments - in one place and assigning roles with drag-and-drop.

Define any role requirements with tags, and auto-fill slots with qualified officers instantly.

Roster's main planning page. Officers can easily search, filter and assign officers to different zones based on their requirements. Auto-assignments are also possible based on matching tags between officers and zone roles.

Every team works differently, so Roster allows teams to configure their roles, skills, zones and requirements differently.

Eliminate ghost assignments and adapt to last-minute plan changes with automatic zone reassignment when an officer is marked absent.

(NOT ACTUAL OFFICER NAMES). Import leave details directly from WhatsApp messages, emails, or notes. The system intelligently extracts names, leave types, and dates from unstructured text.

What we did

We spent the month deeply embedded with ICA teams to understand their actual workflows:

1. Interviewed and shadowed all different teams (A, B, C, D).

2. Ran at least 6 user testing iterations from low-fidelity prototypes to today's working version.

3. Discovered the real struggle faced by officers is assembling scattered information.

Our first user research to understand how ICA officers do roster planning.

Mapping out initial user feedback and prioritising features.

Exploring existing commercial products for inspiration.

Traction

Who tested it

7 officers from all Woodland Teams (A, B, C, D) participated in multiple user testing sessions throughout January. Each team has at least 250 officers and creates rosters daily.

What's happening now

For our pilot, teams will conduct roster planning with both Excel (their current system) and Roster side-by-side. This lets planners verify accuracy and build confidence without operational risk.

The pilot will begin on 30th January with Team C, followed by the other 3 teams.

In our latest user test, an officer assigned 84% in just 40 minutes - putting him on track to complete a full day's roster in under an hour instead of the usual 3-4 hours.

Finished 80% of the roster in 39 minutes in his first try.

What we're targeting

During our pilots, we're tracking whether we've hit these benchmarks

What's uncertain

We won't know the real landed impact until teams start using Roster for actual operations. The shadow pilot will show whether:

• Actual time savings are gained in practice

• Teams trust the system enough to stop using Excel

• The tool handles edge cases we haven't encountered in testing

After User Testing with Team C, Vimal asked if the product is ready to be used.

Team Members

• Edmund (MDDI)

• Fang (OGP)

• Rheza (OGP)

• Si Han (OGP)

• Zong Han (OGP)

Special Thanks

RosterMonster's team for their prior work!

Benjamin, Jing Xuan, and the entire ICA Woodlands crew!