3D Secure 2 challenge
The ACS returns status "C". Your client loads the challenge, and in the local market that increasingly means switching to the issuer's banking app for a biometric approval rather than waiting on an SMS. The customer authenticates there. The app returns control to your checkout and your 3DS Server sends a results request.
Three things in that sequence are outside your control and all three need defensive handling:
-
The app switch itself, which fails silently if the issuer app isn't installed or if deep linking is misconfigured.
-
The return journey, where the customer lands back on a stale page or no page at all.
-
The timing of the final result, which arrives after the customer has already decided your checkout is broken.
Build the return as a server-side resolution rather than a client-side one. The authoritative answer comes from the results request between your 3DS Server and the directory server. Design the checkout so that a customer who returns by any route lands on the real current state of their payment.
Where journeys fail
Abandonment is the baseline. Baymard Institute puts cart abandonment at 70.19% across 49 studies, and authentication friction sits at the most expensive point in that funnel, after the customer has already decided to buy. Every additional second and every app switch subtracts from completion.
The failures that cost more than abandonment are the ambiguous ones. A delayed response that arrives after your timeout leaves two systems disagreeing about the same transaction. An interrupted hand-off produces a 3D Secure authentication the customer completed and your checkout never recorded. A customer who retries creates a second authentication and a reconciliation exception that someone works through by hand.
These compound. 3D Secure 2 challenges that fail on the app switch generate support contacts at a rate out of proportion to their volume, because the customer has no way to tell whether they paid. The support agent can't tell either, unless your logs tie the authentication identifiers to the payment record and the authorization attempt. When they don't, the resolution path is a manual query to the acquirer, measured in days.
Symptoms worth auditing in your own data: payments stuck in an authenticating state past any reasonable timeout and duplicate authentication requests for the same order identifier within a short window. Each one points at a specific defect in how 3D Secure authentication results are persisted and resolved.
Standards and local requirements
Verify against the current specification rather than against your integration's documentation, which ages faster. EMVCo publishes the protocol and the core functions document, and the schemes layer their own mandates on top through Visa and Mastercard program rules. Those are the mandatory controls. Your acquirer will tell you which scheme program version you're certified against.
Locally, the South African Reserve Bank's National Payment System oversight applies, with the Payments Association of South Africa setting industry rules for participants and BankservAfrica operating the clearing infrastructure. The Protection of Personal Information Act governs the cardholder and device data that EMV 3DS messages carry, and the Information Regulator enforces it. Device fingerprinting for 3D Secure authentication is lawful processing, but it still needs a documented basis and a retention policy.
Separate from all of that sits engineering practice, which nobody mandates and which determines whether your checkout works. Idempotency keys on payment creation. Server-side state resolution. Distributed tracing across the authentication and authorization legs with something like OpenTelemetry so a single trace covers both. They appear in the difference between a 94% and an 85% completion rate.
Test the full journey
Certification testing proves your messages are well-formed. It proves nothing about what happens when a customer's phone rings during an app switch. The tests that matter simulate the conditions your production traffic actually meets, and most of them involve something going wrong at the worst moment.
Cover these cases, each with an explicit assertion about the payment's state afterward:
-
Customer abandons at the challenge screen and never returns. The payment resolves to a terminal state within a defined window and no authorization is sent.
-
The ACS responds after your timeout. The late result is reconciled against the existing record instead of creating a second one.
-
The same authentication result arrives twice. The second is ignored and logged.
-
The redirect back to the checkout is interrupted and the customer reopens your site from scratch. They see the current state of the payment.
-
The issuer app isn't installed or the deep link fails. The journey falls back to a browser challenge rather than hanging.
-
Authentication returns "U" or "R". The code follows the documented business rule rather than defaulting to a retry.
-
Authentication succeeds and authorization declines. The customer gets a message that distinguishes this from an authentication failure.
Every one of those tests needs the asserted payment state and the log entries a support agent would use to explain the outcome to a customer. If a test passes but produces no searchable log line tying the authentication identifiers to the order, it hasn't tested enough. Testing 3D Secure authentication properly means testing your observability at the same time.
Review 3D Secure authentication
A useful review starts with evidence rather than architecture. EGS examines your existing flow against your own failure data: the stuck payments and the support tickets grouped by what the customer actually described. That tells us where the breaks are before anyone opens a specification.
From there the review covers the integration boundaries we've mapped above and whether your logs and metrics let you diagnose a failure without contacting your acquirer. It also asks a question that isn't technical: who owns the payment state machine. Flows break at boundaries, and ownership gaps show up at exactly the same boundaries.
The output is a prioritized list of engineering changes tied to EGS's work in card processing and secure transaction engineering, with the expected effect of each on completion and support volume.
Measure engineering outcomes
Checkout completion and support contact volume are the two numbers worth tracking, and both are objectives rather than promises. Instrument them before you change anything, so you have a baseline to argue from. Fixing state resolution and app-switch handling improves completion because it removes failures the protocol never caused. Discuss your journey with EGS and identify the engineering changes needed to address payment friction in your 3D Secure authentication flow.