SYSTEMVALE INDEPENDENT PUBLIC PRODUCT STUDIES
Study date: 12 October 2026

These are independent studies, not client engagements, endorsements or production
improvements. Product names identify the subjects studied.

WHAT IS INCLUDED
- Public NYC 311 responses: a fixed 50,000-row query workload and separate COMPLETE
  monthly aggregates for January and February 2025, with independent total counts.
- One Open Food Facts API response: 100 global breakfast-cereal records, requested
  popularity sort, first page. No representative-catalogue claim.
- GOV.UK Frontend 5.13.0 package and tagged component source.
- Vite issue-form source at dce9fe69d37e957b62f3b5df61a761f7f2f949f5.
- Original local result files, raw benchmark timings, browser checks, workflow
  assertions, a browser screenshot and scripts.
- A manifest with exact source URLs, UTC retrieval times and SHA-256 hashes.

REPRODUCTION (from this extracted folder)
1. Install Python 3.11+.
2. Run: python scripts/analyze-public-studies.py
   No network or third-party Python packages are needed. This regenerates query
   timings, monthly calculations and catalogue decisions, and extracts the pinned
   GOV.UK assets. It uses temporary databases and writes into research/public-studies.
   The published benchmark used Python 3.11.5 and SQLite 3.42.0 on Windows 10.
   Different machines and SQLite versions produce different timings. Compare query
   plans and result equality, not exact milliseconds. Original published results
   are also preserved in the original-results directory.
3. Run: python scripts/study-webhook-reliability.py
   All events and signing material are labelled local test fixtures. No network
   requests or remote actions are performed. The prototype is not a deployment.
4. For the browser experiment, install Node.js and Chrome, run npm install, then:
   node scripts/study-govuk-forms.cjs
   This installs the pinned @playwright/test dependency from package.json.
   The script launches headless Chrome and a temporary loopback server. All form
   submissions stay local. The published run used Chrome 154.0.8037.98.
   The check covers focus, field targeting, returned values, four viewport widths,
   no-JavaScript behaviour and an intentionally broken-target negative control.
   It is not a participant study or WCAG compliance audit.

CAPTURE, NOT LIVE REFRESH
The analysis uses the saved responses. Re-fetching may produce revised data and
must not silently replace the published results. The fetch helper accepts a local
filename and public URL. Some tested endpoints were unavailable: Wikimedia and the
GOV.UK website failed TLS verification from the local runtime; an Open Food Facts
page-100 search returned HTTP 503 twice. They are not evidence for any finding.
The GOV.UK study uses the successfully captured official npm package and GitHub
source. The analytics study uses successfully captured NYC data.

ATTRIBUTION AND LICENSES
NYC 311: NYC Open Data / City of New York.
https://data.cityofnewyork.us/d/erm2-nwe9
This extract contains selected operational fields, not customer names or addresses.

Open Food Facts: Open Food Facts contributors.
https://world.openfoodfacts.org/terms-of-use
Database and derived record-level audit: Open Database License (ODbL) 1.0.
https://opendatacommons.org/licenses/odbl/1-0/
Individual database contents: Database Contents License.
The unmodified selected source response and derived record audit are both supplied.
No product images are included. No dietary or allergen-safety advice is offered.

GOV.UK Frontend: Crown copyright, MIT license. See govuk-license.txt.
Vite: MIT license. See vite-license.txt.
SystemVale's research scripts in this bundle: permission is granted to use, copy,
modify and distribute these scripts with attribution to SystemVale. They are
provided as research examples without warranty; third-party data licenses remain.

CLAIM BOUNDARIES
SQLite: local database timings and correctness, not NYC production or cloud savings.
NYC: request counts and daily rates, not people, satisfaction, causes or forecasts.
GOV.UK: local integration of a pinned component, not a live government service audit.
Open Food Facts: the selected response, not all products or live ranking quality.
GitHub: pinned form review and local synthetic event tests, not live workflow gains.
