For two years, 403 motorists went about Singapore on their daily journeys, just as they always had. Nothing in their cars had changed.

What changed was on the roads around them. A network of 63 cameras was quietly reading their licence plates as they passed, and an app on their phones “charged” them for Electronic Road Pricing (ERP) each time, as if the gantries had never been there at all.

These 403 motorists were our citizen testers, and participating in a live experiment set up by Open Government Products (OGP). They were helping us answer one question: what if there was a different way for ERP to work in Singapore?

Open Government Products (OGP) is an experimental team in government. We build technology for public good, and we build it differently.

In government, execution usually begins only after a plan has been detailed and run through layers of approval. This works well most of the time: we anticipate what could go wrong before it does, and avoid paying for fixes later.

But in technology, things move quickly, and assumptions made during planning don’t always survive contact with the real world. Teams get locked into decisions made before anyone knew better, and by the time something launches, the tech may already be obsolete.

OGP exists because we believe that there is a different way to build. We prefer quick prototypes, we adjust as we go, with citizens involved early, instead of building in isolation and launching all at once.

Each year at OGP’s hackathon, Hack for Public Good, we ask our officers to find real problems affecting citizens and spend the month building towards something that works.

Several of the original team members were drivers or riders themselves, and following announcements that the ERP gantries would be decommissioned, the future of road pricing was on their minds.

Automated Number Plate Recognition (ANPR) cameras were already in use in London, Bangkok, and in major cities across Sweden and Australia. We wondered whether this could also work in Singapore, as an answer to our ageing infrastructure.

LTA was already starting to roll out ERP 2.0, a $556 million project awarded in 2016, and we knew that the solution was unlikely to be dropped.

But that didn’t stop us from asking the question anyway:

We weren’t trying to replace ERP 2.0. We wanted to know if camera-based solutions could actually work in Singapore, as it could matter for whatever comes after ERP 2.0.

There was also the question of whether government could tackle national infrastructure problems differently. We wanted to see if we could use working prototypes to gather real evidence about whether a solution would work, well before we committed to building it out in full.

Here’s how a camera-based road pricing soluction can be implemented in Singapore.

There’s potential for this design to be improved upon if ERP X ever gets a proper nationwide roll-out (e.g. improving the equipment box, using extra SIM cards as a failsafe, using fibre broadband).

But for the purposes of the experiment, this version was already good enough for us to collect meaningful data.

For regular roads, a single camera pointing at 1-2 lanes of traffic works well (97.3% accurate, minimally impacted by nighttime or rain).

But for highways, which are where all 22 remaining active ERP sites are today, the conditions become more demanding. How the cameras are deployed, and how many, makes a real difference to accuracy.

Because we wanted to test how this solution could scale towards a national rollout, we held firm on two principles of camera deployment:

We tested up to 6 lanes each direction, the max in Singapore today. Anything wider might require mounting cameras overhead, which this approach was designed to avoid.

This works out to roughly ~$100K per site per year, and between $38K to $350K per site upfront.

These figures are grounded in actual pilot expenditure, with conservative estimates applied wherever we could not precisely determine them.

For context, we estimate the most expensive ERP X site to cost ~$250K to set up and ~$100K to run in its first year.

Just charging $1 per vehicle and at a poorer 5% leakage rate, that site would still offset its cost within its first year, on a modest 1,009 vehicles per day while ERP is active. For comparison, two lanes of the MCE already see more than 2000 vehicles in a 20-minute window.

While the sample was a self-selected subset of registered ERP X citizen testers and therefore isn’t representative of Singapore’s driving population, the findings suggest that an ANPR-plus-app solution can work well for the people who use it.

The final survey offers valuable insight into what users actually care about: a solution that works seamlessly and reliably, with minimal fuss.

What we found was that involving citizens early, and being transparent with them throughout, actually lowers the risk of unhappiness. Instead of shielding them from the mess of the process and only showing them the polished product, we exposed them to it all.

Citizen testers didn’t mind the scrappiness of the solution, and in fact, seemed to prefer it because it represented a government that was genuine about involving citizens. Many recognised it as a departure from how solutions are usually built, and wanted to see more of it.

Whatever brought people to join as a citizen tester in the first place—whether curiosity, frustration with the status quo, or a hope that camera-based solutions would launch—the process of experimenting together created a shift in perspective.

It had become less about road pricing, and more about what citizen-government collaboration could look like. By the end of the experiment, 96% said they’d want to be involved in future experiments.