As of April 2026
Payment modernization projects do not stall at the architecture layer. They stall at the security review layer. Across multiple major platform transformations, every one lost months to reviews that were not designed for API-first payment infrastructure.
The security teams were not wrong. They were doing exactly what they should. The problem was the process: threat models built for on-prem monoliths, questionnaires written for general-purpose enterprise software, and review cycles calibrated for annual releases.
Where it stalls
Security review
modernization rarely stalls at architecture - it stalls at a review layer built for other software.
Payment-specific risk
Off-template
token-lifetime drift across rails, idempotency replay, and OAuth scope sprawl across four rails.
The fix
Risk register
a joint threat model before design produces one artifact security helped define.
Why the template misses
Payment APIs are not general-purpose enterprise software. They carry threat surfaces a standard questionnaire does not cover: token lifetime drifting across rail boundaries, idempotency replay risk, and OAuth scope sprawl when you integrate four payment rails at once. None of those map cleanly to a generic, NIST-aligned review template, so the review spends its time discovering the shape of the system instead of judging it.
What the fast teams do
They run a joint threat model with security before design starts, not at the end of it. That session produces one artifact: a risk register mapping payment-specific attack vectors to agreed controls, written before a line of code exists. The register then replaces the discovery phase of the formal review, because security already knows what acceptable looks like for a payment API surface - they helped define it.
Payment architects who ship modernization on schedule have one thing in common: a security team that was in the room before the design started.
