← All proof trees

1. 1
   How it was computedvalue = published site capacity summed over the school's own sites (a blank capacity adds nothing; an explicit 0 counts)

   1. Read the committed extract of the district's Facilities Master Plan dashboard (data/cities/philly/reference/parsed/sdp_facilities_sites.csv, one row per site × programme) and keep one row per site.
   2. Link each site to the directory schools (the Master School List) whose ULCS code it carries. A school whose site also hosts another directory school gets no value: that site's figures are the total of every programme on it.
   3. For every other school, combine its own sites: capacity summed (a blank adds nothing, an explicit 0 counts), allotted Pre-K seats summed, and enrollment counted once, because the dashboard repeats the school's whole enrollment on each of its site rows.
   4. Store the school's capacity, its enrollment plus allotted Pre-K seats, its allotted Pre-K seats, and utilization = 100 × (enrollment + Pre-K seats) ÷ capacity. A school with no capacity, a capacity total of 0, or no enrollment gets no value rather than a zero.
2. 2
   The load that wrote it1 data load wrote this metric

   Loaded	Source file	Years	Cells	Suppressed	File hash	Commit
   2026-09-24	sdp_facilities_dashboard	2024-25	223	0	18b9791617ed…	3b369d04

   The hash identifies the exact file version read; the commit identifies the exact code. Both are recorded at load time.

   Evidence: facilities.ts

3. 3
   The publisher's fileFacilities Master Plan dashboard, SY2024-25

   Facilities Master Plan dashboard, SY2024-25 — School District of Philadelphia, Office of Operations. The data behind the district's Facilities Master Plan. Per site: capacity, October 1 enrollment, the Pre-K seats allotted to the site, utilization and the plan's Program Change recommendation. Per building: floor area, year built, condition index, use type and a second capacity figure. One vintage, SY2024-25.

   format

   Qlik Sense app with no file download, read through its Engine JSON API into two committed CSVs

   join key

   ULCS Code; a site can host several programmes and a school can sit on several sites, so the link is a join table

   landing page

   https://www.philasd.org/fpp/

   • Not a file. pipeline_philly/fetch_facilities.py reads the app and writes data/cities/philly/reference/parsed/sdp_facilities_sites.csv and sdp_facilities_buildings.csv; the loader reads the committed site CSV (its hash pin covers both), never the network, and the analysis packet reads the building CSV.
   • Three grains. A site hosting several programmes repeats the same figures on every programme's row (91 of 328 sites). A school on several sites carries its whole enrollment on each of them while capacity is split per site. The building rows repeat a building once per programme on its site (502 rows, 399 buildings).
   • Utilization includes the allotted Pre-K seats (seats, not enrolled children) in the numerator.
   • One vintage: the app's School Year field is cross-joined onto every site and carries no second year of capacity.

   Evidence: source registry (sources.yaml) · sdp_facilities_dashboard: method doc

4. 4
   How it was checkedChecks: 4 passed

   • ✓ passedCheck (iv): each site's published utilization equals (enrollment + allotted Pre-K seats) ÷ capacity on 245 of 245 sites carrying all three, within 0.005; any failure stops the load
   • ✓ passedCheck (v): enrollment with allotment equals enrollment plus allotted Pre-K seats on 285 of 285 sites carrying either; any failure stops the load
   • ✓ passedCheck (vi): a school whose capacity sits on one site reproduces that site's published utilization, 220 of 220 within 0.005; any failure stops the load
   • ✓ passedEnrollment agreement: a school whose own sites carry different enrollments stops the load, since counting it once is right only while the dashboard repeats it; none do in SY2024-25
5. 5
   Assumptions & caveatsAssumptions and caveats

   Assumptions

   • A school on several sites of its own is one school spread across them: the enrollment is counted once and the capacities are added. The dashboard's own per-school measure (Site Utilization (MAP)) follows the same rule.
   • A site shared only with programmes the directory does not carry is attributed to the school it hosts. In SY2024-25 those programmes are all Pre-K (71 of the 223 schools with a value), and the Pre-K seats allotted there are in the school's numerator.

   Caveats

   • SY2024-25 only: the plan has one published vintage, so there is no trend.
   • A school sharing a site with another directory school has no value; its school page shows the site's figures, labelled as the site's.
   • Three schools have capacity on two sites of their own (2210, 7440, 8770). The district's per-site utilization for each of those sites divides the whole school by one site's capacity, so it differs from the school value.
   • Three school values may have a short denominator: Olney Elementary (7400), The Workshop School (8560) and Walter Saul (6040) each have a linked site with a blank or zero published capacity that holds classroom buildings. They are flagged in the method doc and not adjusted.
   • Descriptive, not ranked: the percentile batch leaves out every Facilities key, so these cells carry no peer or citywide percentile.

On a school page this measure is in the enrollment, capacity and utilization section, which is not linked number by number.