How We Evaluate Budgeting Apps: Evidence and Limits

Evidence standard — Current Vault Daily comparisons are based on cited public documentation unless a page provides verified records of an actual test. A plausible scenario, generated dataset or numerical score is not such a record. We are not currently presenting a verified four-week, real-money testing programme.

What a source-based comparison checks

• The exact product, plan, region and billing period being discussed.
• The primary page supporting each material claim, with an access date.
• Whether permissions, costs and limitations are described accurately.
• Whether examples are hypothetical and calculations can be reproduced.
• What remains unknown or could not be independently verified.

What would be required to call something a test

A real test needs a recorded protocol, actual dates, product versions, appropriate consent, traceable observations and a way to reproduce the reported calculation. Identifying details must be protected. We should not claim purchases, account connections, interviews or device trials without records supporting those events.

A test score would also need a documented rubric, transparent weights and underlying observations. We have withdrawn unsupported legacy scores rather than using precise-looking numbers as a substitute for evidence.

The legacy CSV is synthetic

The categorization example explains what the published sample actually contains. It is illustrative synthetic data, not evidence of a measured competition between named apps. It does not validate old rankings.

How AI is used

The publication system uses AI to research, draft, review and publish material through automated checks. A second AI review is not a human fact-check. Source records and deterministic checks help detect mistakes but cannot guarantee truth. See the editorial policy, corrections log and contact page.