Resources
How to run a banking API proof of concept in Australia
Turn an Australian banking API trial into a clear buying decision with representative journeys, acceptance criteria and a practical evidence log.
Fiskil offers a self-service banking API sandbox for evaluating a consented-data integration before production. Its testing guide describes free, unmetered sandbox usage without a credit card or sales call. Use that environment to test your application, then validate real-bank coverage and production access separately.
Fiskil's Banking API has documented linking, data retrieval and sync-event workflows. The proof of concept should establish how those parts fit your release.
Start in the sandbox, then define the production decision
According to Fiskil's testing guide, you can create a Console account, generate credentials and connect a sandbox provider. The guide describes unlimited sandbox end users and consents across Banking and Energy; that is a testing allowance, not a claim about live bank connections or production pricing.
Use a short first run:
- Follow the guide to create the account, credentials and test end user.
- Create an authorisation session and complete the supported sandbox consent journey.
- Wait for the required dataset notification, retrieve the records and finish pagination.
- Test cancellation, a repeated notification and sharing that ends.
- Record what passed in fixtures and what still needs a real-bank test.
Before paying for production, confirm the banks, account types and data your app needs, how you can access them, and the price and support included. Fiskil's go-live checklist describes the separate production-readiness process. A sandbox success is a useful engineering result; keep the production acceptance decision explicit.
Write a one-page test charter
Choose one real product decision. A spending app needs a coherent transaction view; an onboarding product may need account information for a matching process. Those are different tests even when both start with “connect a bank”.
Record the launch customer, required institutions, account products, datasets, intended use and success conditions. Separate requirements needed at launch from later expansion. Ask who approves the evidence: engineering can establish integration behaviour, while product and the responsible business owners assess the customer journey and access arrangement.
The following is a proposed charter, not a claim about every supported connection.
| Test | Evidence to collect | Decision it supports |
|---|---|---|
| Customer approves sharing | Selected accounts and consent outcome | The intended journey can complete |
| Transactions become available | Dataset notification, retrieval and pagination result | The feature can load its required history |
| Customer declines | Return path and visible next step | An incomplete connection is handled |
| Sharing ends | Backend and interface behaviour | The feature respects its access lifecycle |
| One institution is unavailable | Scoped error and recovery route | One failure does not confuse every account |
Separate sandbox and production evidence
Fiskil's testing guide provides sample-data journeys across banking and energy. Use them to test your backend, interface and event processing without treating a sandbox result as proof of live bank coverage.
For example, a synthetic test might require an application to display two accounts, process a repeated notification once and finish all transaction pages. Passing that test shows the application handled those fixtures. It does not establish that a particular business banking portal is available in production.
Keep a second list for live validation: institution, customer eligibility, banking channel, required fields and observed data availability. Mark untested cases explicitly.
Follow the data beyond the redirect
The banking guide describes separate sync events after account linking. A completed browser journey and a complete application view are different milestones.
Measure each stage against your own acceptance policy. Record the notification received, the retrieval outcome, any missing pages and when your own processing completed. Distinguish an empty dataset from an error.
Include a support handover: can someone identify the affected customer and request without copying full banking records into general logs? Can they explain what the user should do next?
Make the buying decision reviewable
Finish with a short evidence log: passed, failed, untested and accepted with a limitation. Attach unresolved coverage questions and the agreed operating responsibilities. Avoid combining every result into a score that hides a failed mandatory requirement.
Fiskil's production checklist gives the next set of readiness questions. Discuss your test charter with Fiskil, then choose against demonstrated requirements rather than the appearance of the first demo.


