Key takeaways
- Measure against a recorded promise or policy interval.
- Test message support separately from message timing.
- Minimize resident information in the study dataset.
Source record
3 cited sources
Last verified
2026-09-14
Fast repair messages are not necessarily useful, and useful messages can still arrive after a promised time. This design measures timeliness and factual support as separate outcomes.
Research question
Among eligible open repair cases, what share of scheduled or promised resident updates are delivered by the recorded deadline, and what share accurately reflect the latest confirmed work record?
Methodology
Define eligible cases, the event that creates an update obligation, and the timezone used for deadlines. Build a promise-level table with case identifier, obligation source, due time, actual delivery time, approved channel, and delivery evidence. Treat an absent promise separately from a missed promise.
For a sampled content review, compare each message with the latest confirmed vendor, access, approval, and work-order records available before it was sent. Code supported, partly supported, unsupported, or indeterminate. Use a second reviewer for borderline cases and retain disagreement counts.
Evidence
The NIST Privacy Framework offers a structure for managing privacy risk. HUD tenant resources provide general housing context. The W3C provenance recommendation provides a vocabulary for describing the relationship between records and activities.
Source treatment
Keep message bodies in the approved communication system. Use pseudonymous case identifiers in the analysis table and exclude unnecessary names, contact details, and access information. Record delivery evidence separately from a platform's sent timestamp when both are available. Never infer receipt from a drafted message.
Limitations
Channel receipts do not always prove that a resident read a message. Emergency cases may follow different rules and should be reported separately. Missing historical promises can bias the rate upward. Content coding tests support in the available record, not whether every statement was ultimately correct.
Interpretation
Report the timeliness denominator, excluded cases, and content sample size. Examine late and unsupported updates by workflow stage before assigning responsibility.