Build log
What we're building in public
Placet is built openly. Here's what's shipped, what's in flight, and what's next.
Shipped
11 itemsAcute Rehab (IRF) — 1,221 inpatient rehabilitation facilities now searchable
Inpatient Rehabilitation Facilities from CMS IRF Compare are now searchable via the Acute Rehab tab on the facility finder. Quality metrics include readmission rate from CMS IRF Compare data.
LTC expansion: ALF, CCRC, and memory care facility types
509 NC ALFs and 531 NY ALFs ingested alongside 39 CCRCs. New facility_type column on the facilities table with constraint-checked values. Searchable via new Assisted Living and CCRC tabs.
CCRC financial profiles
Contract types (life-care, rental, fee-for-service), entrance fee ranges, monthly fee ranges, and unit counts added for CCRC facilities.
Facility comparison expanded to 20 facilities
Compare modal now handles up to 20 facilities simultaneously with horizontal scroll, so social workers can compare a full referral list in one view.
Semantic color coding for quality metrics
Quality measure cells now render green/amber/red based on clinical thresholds rather than raw values, making risk legible at a glance without explanation.
Medicaid acceptance as primary payer signal
Medicaid acceptance surfaced as a first-class filter and badge. Addresses the referral platform industry's systematic deprioritization of non-commission Medicaid placements.
Referral inquiry emails now actually send
Referral inquiry notifications were previously mocked. Now delivered via Resend with facility context, admissions contacts, and direct inbox links.
Named admissions contacts
Social workers and intake liaisons now stored per facility. Referral inquiries route to named contacts rather than generic facility inboxes.
Trust Index v2 with Four-Pillar Trust Bar
Rebuilt the Trust Index from a single composite into four independent signals — RN hours, CMS stars, Trust Index, and bounce-back rate — each with its own color-banded display.
RN hours clock visualization
Clock-face graphic that makes staffing hours tangible: how many minutes of RN time per resident per day, at a glance.
For-profit / low-census penalty removal
Removed the automatic Trust Index penalty for for-profit ownership and low census after determining the signal was too noisy to be fair at the facility level.
In flight
5 itemsNPPES + CMS data ingest crons (monthly / weekly / quarterly)
Automated pipelines to pull NPI bulk downloads and CMS provider/staffing/quality datasets on their respective release schedules.
Ownership network graph — entity resolution
Building normalize_operator_name, a Postgres function that collapses variant legal strings (LLC vs L.L.C., abbreviations, punctuation drift) into canonical operator identities.
ALF and CCRC data expansion beyond NC/NY
State licensing data coverage is patchy. Expanding ALF ingestion to additional states and working on a standardized pipeline for state-regulated facilities that lack CMS CCNs.
PA PHC4 bounce-back data request
Formal data request to Pennsylvania's Health Care Cost Containment Council for facility-level readmission data to supplement CMS claims.
LTCFocus longitudinal data integration
Mapping Brown University's LTCFocus annual SNF panel data into our schema to enable year-over-year quality trend views.
Next up
4 itemsSplink probabilistic matching for ownership dedupe
Apply the UK Ministry of Justice's open-source probabilistic record linkage library to match operator entities across CMS, NPPES, and state enrollment records.
Human review queue UI for ambiguous operator matches
Internal tool for flagging and resolving entity pairs where automated confidence scores fall below threshold — keeping a human in the loop.
SEC CIK map for public chains
Cross-reference publicly traded nursing home chains against SEC EDGAR CIK identifiers to surface 10-K filings and ownership disclosures.
State-level data layering
Pull and normalize state Department of Health inspection and complaint records beyond the seven states currently ingested.
Tell us what to build
Suggest a feature
What would make this tool more useful for you? What data are you missing? What's frustrating?