Start typing to search this publication.
techcirkle
Open menu

Subscribe to techcirkle

Get new posts delivered straight to your inbox.

Singpass, PayNow and the Integration Timeline Nobody Puts in the Estimate

The engineering for Singapore's national integrations takes days. Getting access takes weeks. Here is how that reshapes a project plan.

techcirkle
5 min read
Cover image for Singpass, PayNow and the Integration Timeline Nobody Puts in the Estimate

An engineer looks at the Myinfo API documentation, sees an OAuth flow returning a JSON payload of verified personal attributes, and estimates three days. The engineer is correct. Implementing Myinfo takes about three days.

Getting permission to call it in production takes considerably longer, and that number is the one that belongs in your project plan.

Singapore business district skyline
Singapore business district skyline

This is the most consistent scheduling error I see on Singapore builds. Teams estimate national integrations by code complexity, discover the access process in month three, and lose a month of runway to paperwork that could have been started on day one. The engineering is genuinely easy. The critical path is administrative.

What actually gates each one

Singpass and Myinfo. You need a registered entity, a stated and justified purpose for each attribute you request, and — for the production tier — a review of how you handle the data you receive. Sandbox access comes quickly; production does not. Expect the end-to-end process to run several weeks, longer if your purpose statement requires clarification. Requesting more attributes than you can justify is a common cause of delay: ask for the minimum set, because every extra field invites a question.

PayNow and FAST. These are bank-mediated. You are not integrating with PayNow directly; you are integrating with a bank or a payment service provider that reaches those rails. That means bank onboarding, which means KYC on your entity, a commercial agreement, technical certification, and — depending on your model — a discussion about whether you are handling funds in a way that carries licensing implications. Plan in months if you are going direct to a bank; weeks if you are going through an established PSP.

GIRO. For recurring collection, and it is the sleeper. Mandate setup involves a paper or digital authorisation flow, bank-specific file formats, and a settlement cycle that is not instant. Products designed around card-style immediate confirmation need reworking once GIRO's timing is understood.

Corppass if you serve business entities rather than individuals — a separate registration path from Singpass with its own timeline.

What this means for sequencing

The rule is simple and almost never followed: start every external access application in the first two weeks of the project, before the code that consumes it exists.

This feels wrong to engineering teams because you are applying for access to something you have not built yet. Do it anyway. The applications are cheap to submit and the waiting is the expensive part. While approvals process, build against your own mock implementations of the same contracts.

That mock layer is worth building properly rather than as throwaway scaffolding:

interface IdentityProvider {
  authorize(scopes: Attribute[]): Promise<AuthCode>
  fetchProfile(code: AuthCode): Promise<VerifiedProfile>
}

// MyinfoProvider   — real, behind approval
// MockProvider     — deterministic fixtures, always available
// SandboxProvider  — the government sandbox, available early

Three implementations of one interface. Development proceeds against the mock, integration testing moves to the sandbox as soon as it is granted, and production swaps the binding. You have also, incidentally, built the thing that lets you test error paths — which the sandbox will not reliably reproduce for you.

Two engineers reviewing integration code
Two engineers reviewing integration code

The error paths you have to design for

Sandbox environments show you the happy path. Production shows you everything else, and these are the cases that generate support load:

  • Myinfo returns partial data. Not every citizen has every attribute populated. A registration flow that assumes a complete profile will break on real users. Design every field as optional with a manual fallback.

  • The user abandons mid-consent. They land on the Singpass consent screen and close the tab. Your session needs to expire cleanly and be resumable without duplicating the record.

  • Attribute mismatch. The verified name from Myinfo does not match what the user typed, or what your existing record holds. This needs a defined resolution path, not an exception page.

  • PayNow proxy resolution fails. A mobile number or UEN that does not resolve to an account. The error surfaces late in the flow and needs to be handled with a clear message rather than a generic failure.

  • Settlement timing. Instant transfer is instant most of the time. Your reconciliation must tolerate the times it is not, without double-crediting.

Each of these is straightforward to handle if you design for it. Each is a production incident if you do not.

Budget consequence

The integration work itself is a small share of a project. The lead times are not, and they interact badly with fixed-date commitments.

If you have promised a launch date to a board or an investor, the honest thing to do is work backwards from the longest approval timeline in your dependency set rather than from your engineering estimate. In practice that usually means the bank relationship, and it usually means starting it before you have finished deciding what you are building.

For the wider picture — realistic cost bands, contract structures, PDPA and MAS TRM architecture, and a 90-day plan for a Singapore build — see the full guide: Application Development Singapore: What Senior Buyers Get Wrong in 2026.

TechCirkle builds these integrations regularly as part of custom software development work in Singapore and ASEAN.

More from TechCirkle

Frequently Asked Questions

How long does Singpass Myinfo integration take?

The code is roughly three days. Production access typically takes several weeks, covering entity registration, purpose justification for each attribute requested, and a review of your data handling. Sandbox access is much faster. Request the minimum attribute set — every additional field invites review questions and adds delay.

Can I integrate with PayNow directly?

Not as a typical business. PayNow and FAST are bank-mediated rails, so you integrate through a bank or a payment service provider. Going direct to a bank means KYC, a commercial agreement and technical certification — plan in months. An established PSP compresses that to weeks but adds per-transaction cost.

What is the most commonly underestimated Singapore integration?

GIRO. Recurring collection involves mandate authorisation flows, bank-specific file formats and a non-instant settlement cycle. Products designed around card-style immediate confirmation usually need rework once GIRO's timing characteristics are properly understood.

How should I structure code around approvals I do not have yet?

Define the integration as an interface with three implementations: a deterministic mock, a sandbox provider, and the production provider. Build against the mock from day one, move to sandbox when granted, and swap the binding for production. The mock also gives you reliable error-path testing that sandboxes rarely provide.

What error cases matter most in Myinfo flows?

Partial profiles where attributes are not populated, abandoned consent mid-flow, and mismatches between verified attributes and existing records. Treat every field as optional with a manual fallback and define an explicit resolution path for mismatches rather than surfacing an error page.

Do I need Corppass as well as Singpass?

If your product serves business entities rather than individuals, yes. Corppass handles corporate digital identity and has its own registration path and timeline, separate from Singpass. Products serving both need both, and they should be scoped as two integrations rather than one.

When should I start the approval applications?

In the first two weeks of the project, before the consuming code exists. It feels premature, but the submission is cheap and the waiting is what costs you. Work your launch date backwards from the longest approval timeline in your dependency set, not from your engineering estimate.

Subscribe to techcirkle