A production smoke test should answer whether the deployed core path is working now. Keep it small enough to run consistently after releases and safe enough that it does not create real customer messages, charges, or operational commitments.
Choose the path from the product's most important promise. A homepage returning successfully is weak evidence when customers depend on authenticated writes and private retrieval.
Define a controlled account and fixture
Use an authorised test account with clearly labelled fictional data and bounded permissions. Decide which actions are safe in production, which are simulated elsewhere, and how test records are identified and removed under the data policy.
Record the deployed version and environment at the start. That prevents a successful check against a stale preview or old process from being mistaken for evidence about the new release.
Smoke-test a fictional study workflow
For a study product, a focused check might sign in, create a controlled draft, save a question, reload it, verify private access from a second account is denied, and remove the fixture through the supported lifecycle.
If publishing or respondent submission is part of the safe fixture design, exercise one controlled response and confirm it appears in the right workspace. Keep outbound notifications directed only to authorised test recipients. Do not include live payment attempts merely to make the check look comprehensive.
Verify the stored result and a safe diagnostic reference. A screen showing success while storage failed is exactly the kind of deployed defect the smoke test should catch.
Keep deeper tests outside the smoke path
Concurrency, restore exercises, broad permission matrices, and payment lifecycle permutations belong in their appropriate verification processes. Repeating all of them after every small release can make the smoke test slow enough that people skip it.
Add a new smoke assertion only when it protects a critical deployed boundary and remains safe to execute. When a production incident reveals a missing check, decide whether it belongs in the smoke path or a more targeted regression test.
The tradeoff is breadth versus reliability of execution. A short meaningful check run every time can provide more practical value than a broad checklist completed inconsistently. It still does not prove overall correctness or future availability.
The current Next.js production guide supplies framework context. The StackFlight case study provides a connected study-workflow reference, while this smoke test is a proposed exercise rather than a reported procedure there.
Questions about production smoke tests
Is checking the homepage enough?
Only for a product whose promise ends there. Most SaaS pilots need at least one authenticated data path verified.
Should test data remain permanently?
Use a deliberate fixture and cleanup policy. Avoid uncontrolled accumulation or records that contaminate customer reporting.
What should the test report contain?
Record version, environment, safe account, executed path, stored outcome, cleanup, and failures. Bring unclear boundaries to a release-process review.