Test stacked health plan filters and what they exclude

The checks that matter

Each stage has a count, and the counts are what make the filtering checkable rather than merely plausible.

The application exposes an excluded list, and asserting it is the rare chance to check the hidden half of a filter directly. A plan wrongly excluded is a lost sale that nobody can see from the results.

The flow as a test file

The surviving plans are asserted with their premiums, which proves the premium filter was applied to the right field rather than coincidentally matching.

---
mode: testing
url: https://my-testing-repo-main.vercel.app/policybazaar-clone-app/health?reset=true
max_steps: 40
tags: [policybazaar, insurance, filters]
---

# Policybaazar 27.3: Health plan filter

Catalog objective: filter health plans by room rent limit and premium.
Key assertion: the results respect both filters.

## Verify the unfiltered list
Verify "Plans shown" reads "5 of 5".

## Apply the room rent filter
Select "No room rent cap" in "Room rent" and verify "Plans shown" reads "3 of 5" listing "Meridian Care 10L", "Harbour Total 20L" and "Meridian Essential 7.5L".

## Add the premium filter
Type "15000" into "Maximum annual premium" and verify "Plans shown" now reads "2 of 5".

## Verify both filters were applied
Verify the results are "Meridian Care 10L" at "₹14,280.00" and "Meridian Essential 7.5L" at "₹11,430.00", and that "Harbour Total 20L" appears in the "Excluded by these filters" list.

The two filters are applied in separate steps so the count can be checked between them, localising a fault to whichever facet narrowed the set wrongly.

Step by step coverage

A baseline and three progressively narrower checks.

Naming the excluded plan is what turns a filter test into a check on the rule rather than on the result count.

Adding it to a suite

kane-cli testmd run tests/policybazaar/health-plan-filter_test.md

Add a combination that excludes everything, since an empty state on a comparison site is a moment to offer alternatives rather than a blank panel.

Filters read most like the requirement they came from. From a PRD to evidence covers that alignment, and generating filter combinations covers producing the set.