ElderGrade.com

The problem

The data families need before choosing a nursing home is already public. CMS publishes inspection citations, payroll-based staffing, staff turnover, enforcement actions and full ownership records for all 14,693 certified homes in America.

Almost nobody can use it. The files run to hundreds of thousands of rows with cryptic column names, and the most decision-relevant fields are buried where no family will ever find them. Meanwhile the sites families do find are lead-generation businesses paid a referral fee on placement — so the data they most need is precisely the data those sites have no reason to show.

What we built

We built a reader for the CMS Provider Data Catalog that takes field names from the JSON API and row data from the CSV export, and refuses to import at all if the two drift apart — so a schema change upstream fails loudly instead of silently corrupting the site.

On top of that sits an original grading engine: 100 points across inspection safety, staffing, staff stability, resident outcomes and enforcement history. Staffing thresholds are percentiles recomputed from the live population on every import, so the curve moves when CMS republishes. Three findings cap a grade outright, and a home with too little published data is left ungraded rather than given a flattering default.

Every component, deduction and cap is stored on the record and rendered on the page, so any reader can reproduce the arithmetic.

Where it landed

A site that publishes what the incumbents hide: the inspector's own words on every citation, staff turnover, the weekend staffing gap, and the ownership chain behind each home. Families can also review a home, and facilities can claim their profile and reply — with a hard firewall that keeps both out of the grade, because a grade that reviews could move would be a grade that could be bought.

62,000 indexable URLs, built with programmatic pages capped at a threshold so thin pages never enter the sitemap.