How DataVault™ Computes Honest Numbers
CardSahayak never invents metrics, never prints unverified percentages from tiny sample sizes, and never exposes individual user identities. Every community metric passes through six mathematical safeguards.
Official Bank SSOT Baseline + Community Telemetry
Every card page displays the bank's official MITC parameters (minimum income, CIBIL score cutoff, annual fee waiver spend threshold, and quarterly lounge spend rule) alongside community-submitted outcomes. If a bank does not publish a CIBIL cutoff, we explicitly state "Not Published by Bank" rather than assuming an arbitrary figure.
Groups With Fewer Than 5 Reports Are Hidden
Any card or demographic slice with fewer than 5 verified reports (`n < 5`) is withheld from percentage display. Furthermore, if exactly one demographic slice falls below 5 reports, our engine automatically applies complementary suppression to the next-smallest visible slice so the hidden group cannot be back-calculated from the total.
Counts Instead of Raw Percentages Below 30 Reports
When a group has between 5 and 29 reports, DataVault prints exact counts (e.g., 8 of 10 approved) rather than a raw percentage, preventing small samples from masquerading as high-precision rates.
Every Proportion Includes Its Likely Range
Alongside every visible rate, we compute the two-sided 95% Wilson score confidence interval (z = 1.96). For example, 8 of 10 approvals shows Likely 49–94% (Early Signal), whereas 800 of 1,000 approvals shows Likely 77–82% (High Confidence).