## TL;DR

Cypress Cloud detects flaky tests by watching retries: a test that fails then passes on retry gets flagged. It is zero-effort but costs money, keeps your data in their system, and uses their definition of flaky. Custom tooling (parsing JUnit XML or your CI API into your own flake scores) costs engineering time but gives you your own thresholds, your own history, and alerts in your own channels. Small teams should start with Cloud; teams with strong data practices or cost pressure should build.

## The query

```text
cypress cloud flaky detection vs custom tooling
```

## Use this when

- Deciding whether to buy Cypress Cloud for flaky detection.
- Scoping an internal flaky-test dashboard project.

## Not for

- Fixing a specific flaky test.
- Non-Cypress suites; the Cloud half does not apply.

## Steps

1. List what you need: detection only, or also quarantine, alerts, and history. Expected output: a short requirements list.
2. Trial Cypress Cloud detection on your real suite for two weeks. Expected output: a list of flagged tests to sanity-check.
3. Prototype custom scoring on the same two weeks of JUnit data. Expected output: your own flake ranking to compare against Cloud's.
4. Compare the lists plus cost and maintenance effort honestly. Expected output: a decision backed by data, not marketing.
5. Revisit yearly; suite size and team shape change the answer. Expected output: a scheduled re-evaluation, not a forever decision.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_zPESH0vDnrNbO9yF5N6_pA