Client engagement · Alliance Against Seclusion & Restraint
A public dataset people can explore.
AASR needed to make federal school restraint-and-seclusion records understandable. Prior Signal developed a data experience and supporting processing and methodology work.
The brief
Make the record legible without overstating it.
The U.S. Department of Education publishes Civil Rights Data Collection records. The project translated those files into a resource for people who need to examine the data without starting from raw spreadsheets.
The work joined data processing, editorial structure and implementation. It does not establish that the underlying administrative records capture every event, nor does matching selected official totals validate every possible interpretation.
Civil Rights Data Collection — primary sourceThe output
What the engagement produced.
Data experience
A self-contained browser artifact designed to make the underlying records easier to explore.
Processing workflow
A Python pipeline and selected headline checks against official publications, with discrepancies identified for review.
Supporting datasets
Structured exports and documentation of fields, suppression codes and transformations.
Methods and handover
Documentation explaining source releases, calculations and the limits of interpretation.
Accessible alternatives
Data-table alternatives and keyboard and reduced-motion considerations alongside the visual presentation.
Publishing materials
Static-reference and share-card work supporting distribution of the project.
These are deliverables, not an independently audited return-on-investment claim. This case study does not quantify staff time saved, public use or downstream policy outcomes, and it does not present a checklist score as a comprehensive accessibility certification.
Corrections and interpretation
What an earlier version got wrong.
Representation is not relative risk.
An earlier version labeled an enrollment-share comparison as a risk ratio against non-disabled peers. That wording and numerical risk claim have been removed. A valid relative-risk calculation requires matching affected-person counts and enrollment denominators for both groups. We are not substituting a new ratio without that calculation.
No positive count is not the same as a missing record.
An earlier version treated schools without a positive restraint-or-seclusion count as having filed no usable data. That conclusion does not follow. Valid zeros, missing values, suppressed values and invalid reports must be distinguished. GAO has documented inaccurate zero reporting, but that does not make every zero a missing value.
GAO — reporting zeros and missing dataComparisons require matched definitions and releases.
People affected and repeat incidents are different measures. IDEA-only and combined disability categories are also different populations. A cross-year comparison must name the source releases, definitions and exclusions. Specific delivery-year descriptions on this page have been removed until reconciled against the delivered release manifest; the methodological article separately explains the comparisons it discusses.
The externally hosted data experience is a separate artifact. Changes to this portfolio page do not automatically update its figures or certify those figures as correct.
Read the methodological articleA bounded engagement
Have records people need to understand?
Start with the intended audience, the source material and the questions the final resource must answer. Agree the output, checking process and handover before the build begins.