Work / Singapore Customs

Singapore Customs · Information architecture & content

402 pages, restructured around what users came to do.

A high-traffic government website serving 3.9 million visits a year. The content was accurate and well maintained. What was broken was how anyone found it, especially on mobile, and especially for first-time visitors who were not even sure they were on the right agency's site.

Role
IA and content restructure lead
Client
Singapore Customs
Scope
Full IA restructure, content rewrite, mobile rebuild
The shift
Structure built around the org chart, rebuilt around user tasks
Illustration: a tiny Xiaohei with a map dwarfed by a towering wall of 402 department-labelled boxes, unsure which to open, captioned where. Xiaohei illustration style by Ian.

At a glance

The problem
3.9 million visits a year, but the site's search failed every time and first-time visitors gave up before finding what they came for.
The move
Rebuilt 402 pages: from a site sorted by who you are to one sorted by what you need.
The outcome
First-time user confidence rose 4.2 to 8.0. The group that had struggled most improved most.

About the project

The content was not the problem.

The agency knew its subject matter. What had accumulated over the years was structure shaped by the organisation chart: content grouped by the department that owned it, not by the task a visitor arrived to complete.

There was a complication we could not design away. Many visitors were not sure they were on the right government site at all, confusing this agency with a related one. Fixing the brand confusion was out of scope. So the navigation had to be clear enough that people could still reach what they needed even while uncertain where they had landed.

The challenge

Users of government platforms tolerate friction quietly.

Professionals who used the site daily had built workarounds. First-time visitors gave up before finding what they came for. Almost none of it was reported, because people do not complain to a government website; they route around it. The research had to actively look for friction users had stopped mentioning.

The process

Three groups, two directions, one buildable answer.

01

Research across every group the site serves

Before touching a single page, I scoped research across three distinct groups: individual members of the public, external business users, and the internal officers who process what the site generates. The findings put numbers to the pain: in testing, the site's own search sent people away empty-handed every single time, a hundred percent abandonment, and first-time visitors rated their confidence in finding anything near the floor.

31 participants 3 user groups In-depth interviews Focus groups 100% search abandonment
02

Auditing all 402 pages

Every page was mapped, assessed, and assigned a fate: keep, rewrite, merge, or retire. The audit revealed the core pattern: structure mirrored departments, while users arrived with tasks. The same task often crossed three departmental sections.

402 pages audited Content inventory Task mapping
03

Two directions, then a third

Rather than jump to one answer, I built two. Version A was radically user-centric, organised entirely around tasks. Version B was a safer evolution that kept the organisation's existing shape. I documented the trade-offs of each, user impact against implementation complexity, and pressure-tested the task-based model against how global customs authorities structure the same content, from Australia to US Customs and Border Protection to New Zealand. The client asked for a blend of the two, and that became Version C: the user-first spine of A, made buildable by the pragmatism of B.

Version A vs Version B Trade-offs documented Benchmarked: AU, US CBP, NZ Version C, the blend
04

Building Version C, written for action

The new architecture regrouped everything into task-based actions. Content was rewritten throughout to prioritise what users needed to do, not what the organisation needed to say. Mobile was rebuilt as a primary surface rather than an afterthought, within the constraints of the whole-of-government platform the site runs on, which meant I negotiated what the system allowed rather than designing in the abstract.

Task-based IA Full content rewrite Mobile-first rebuild Platform constraint negotiation
05

Validating with the same people, after launch

Post-revamp validation went back to all three user groups. The sharpest movement came from the group that had been struggling most: first-time individual users.

Post-launch validation Same 3 groups
Illustration: a first-time Xiaohei pulling the right answer card out of a phone in one confident move, labelled first try and found it. Xiaohei illustration style by Ian.

The contested moment

Arguing against the org chart.

Departmental structure is comfortable for the organisation: every team can see its own content. Task-based structure is comfortable for users and uncomfortable for everyone who loses a dedicated section. I held that position by showing each stakeholder, with their own users' research in hand, that being findable matters more than being visible.

Results

Validated after launch, across every group.

4.2 → 8.0
first-time individual user confidence, the group that had been struggling most
5.1 → 7.8
ease of use for the same first-time users
6.3 → 7.8
overall confidence across all three user groups

The brand confusion that sent people to the wrong agency's site in the first place remains unresolved; it was outside our scope and still is. The honest claim is narrower: when people land here, even uncertain, they can now find what they came for.

What stayed with me

Government platform users do not complain. They build workarounds and come back and try again. The real design work is actively looking for the friction people have stopped reporting.
Xinning Chia, IA and content restructure lead

Other cases

More decisions, told the same way.