Methodology
Every Tekify directory follows the same pipeline: download the source dataset, clean it,
join supporting data, render one page per record, and state the source. This page documents that pipeline
and its known limitations.
Pipeline
- Source. Each directory has one primary dataset published by a public agency or an
open-licence project (US Census, CMS, NCES/IPEDS, GeoNames, OurAirports, Razorpay IFSC).
- Parsing. The file is parsed by code, with encoding and column-name handling written
for that specific dataset. Record identifiers (ZIP, IFSC, CCN, unitid, IATA) are treated as strings, not
numbers, so leading zeros survive.
- Cleaning. Whitespace and casing are normalised for display; missing values are marked
as unavailable rather than guessed or imputed.
- Joins. Where a page needs extra context — coordinates for nearby-record links, county
assignment for places, income bands for colleges — the record is joined to the supporting dataset on its
identifier. Unmatched records are kept and simply omit the joined fields.
- Rendering. One page per record with the fields that matter, the surrounding records
(same state, city, bank or rating band) and the source statement.
- Validation. Every build checks for duplicate titles and descriptions, broken internal
links and empty pages before deployment.
Ratings
Star ratings (hospitals, nursing homes, home-health agencies, dialysis facilities) are reproduced
exactly as CMS publishes them. They are not recalculated, reweighted or adjusted. When CMS marks a
facility unrated, the page says so — unrated is not a poor rating.
Updates
| Dataset | Source cadence | What a refresh changes |
|---|
| CMS quality data | quarterly | ratings, measures, staffing figures |
| Census population estimates | annual | population figures |
| IPEDS college data | annual | admissions, tuition, net price |
| OurAirports / GeoNames | continuous | airport and place records |
| Razorpay IFSC | continuous | bank branch records |
Known limitations
- Derived figures (distance between records, state averages, percentages) are computed from the data as
published and inherit its errors.
- Records can lag the source by the time between a source update and the next import.
- Some source fields are optional and blank for many records; pages show "—" rather than a guess.
- Pages are static: a record changed at the source today appears after the next import.
Verifying a record
Each page names its source dataset. To verify a figure, compare it with the agency's own lookup tool —
for example CMS Care Compare for hospital ratings, or the Census Bureau for population. If you find a
discrepancy, the contact page on tekify.xyz is the fastest way to
report it.