Vacancy Utility Charge Leakage Study

Published October 13, 2026

Research Question

Vacancy Utility Charge Leakage Study examines this question: Which charge periods fall outside documented occupancy and transfer windows, and what source gaps prevent administrative resolution? The unit of analysis is one utility invoice line for one property-meter billing period. Defining that unit before extracting records prevents property-level, account-level, and event-level observations from being mixed in one denominator.

The working hypothesis is that Meter-level timelines will reveal recurring transition gaps that portfolio totals conceal. The hypothesis is directional, not a claim that the relationship has already been proven. A portfolio should test it on its own authorized records and retain contrary results.

Evidence Model

The evidence model separates source facts, administrative actions, accountable decisions, and observed outcomes. Source facts reproduce what an authoritative record contained at a defined time. Actions show what staff routed or completed. Decisions identify who exercised authority. Outcomes document the later state. A reliable analysis never infers a decision merely because an action occurred.

Six dimensions organize the review:

• Meter Identity: record the source, observation rule, missing-value treatment, and reviewer.
• Billing Window: record the source, observation rule, missing-value treatment, and reviewer.
• Possession Timeline: record the source, observation rule, missing-value treatment, and reviewer.
• Transfer Confirmation: record the source, observation rule, missing-value treatment, and reviewer.
• Charge Allocation: record the source, observation rule, missing-value treatment, and reviewer.
• Resolution State: record the source, observation rule, missing-value treatment, and reviewer.

Missing values need their own codes. Not collected, unavailable, not applicable, access restricted, and expected later are analytically different. Converting each to blank or zero creates an apparently complete dataset that cannot support an honest conclusion.

Proposed Method

The proposed method is to link invoices, meter identifiers, move events, and provider confirmations for a frozen cohort; classify overlaps and retain unresolved responsibility. Define the observation window, inclusion rule, exclusion rule, and cutoff before reviewing outcomes. Keep a frozen raw extract beside the transformed analysis table, with a data dictionary that explains every derived field.

Build a stratified sample that includes ordinary cases, aged open items, recently corrected records, and exceptions. Random sampling of only completed work can hide failures that remain stuck upstream. Where records are clustered by property or provider, preserve that relationship rather than treating every observation as independent.

Two reviewers should code a pilot independently. Compare their classifications, discuss disagreements, and revise the definitions before expanding the sample. Agreement does not prove that the source is correct, but disagreement identifies ambiguous instructions or evidence that cannot be interpreted consistently.

Measurement Plan

Report counts and rates with their denominators. For duration, show medians and relevant percentiles alongside the oldest open cases; an average alone can conceal a small group of consequential delays. Separate active handling time from external waiting when the timestamps support that distinction. Preserve open cases as open rather than assigning them an artificial completion date.

Pair outcome measures with evidence-quality measures. A fast process with weak source linkage is not necessarily reliable, while a slower process may reflect explicit approval or access controls. Suggested measures include dimension completeness, exception rate by reason, reviewer agreement, reopened records, age by current state, and the share of items with a reproducible closure test.

Do not create a composite score until each component has a clear operational meaning. A single number can conceal whether improvement came from better evidence, fewer hard cases, changed definitions, or excluded records. A small dashboard of interpretable measures is more useful for management decisions.

Bias, Confounding, and Data Quality

Administrative records are produced by a workflow, not by a controlled experiment. Staff may backfill timestamps, difficult cases may generate more notes, and system changes may alter what is captured. Missing evidence can indicate workload, unclear ownership, restricted access, or an event that occurred outside the primary tool. Describe those possibilities rather than treating missingness as random.

Comparisons can also be confounded by property type, jurisdiction, service scope, season, vendor market, portfolio policy, or resident circumstances. Stratification can make patterns more legible, but small groups should not be reported in ways that expose individuals. Any observed association remains a prompt for further review, not proof of causation.

Test sensitivity by changing reasonable definitions: the start event, the closure rule, the age bands, or the handling of corrected records. A conclusion that disappears under a small defensible change should be reported as fragile. Preserve both specifications and explain why one is preferred.

Governance and Privacy

Use only records the review team is authorized to access. Minimize resident, owner, employee, vendor, financial, and account data in the analysis table. Replace direct identifiers with controlled study IDs, store the re-identification key separately, and set a retention date. Published summaries should suppress small groups or unique narratives that could reveal a person or property.

Assign a study owner, data custodian, method reviewer, and decision owner. Record changes to definitions with an effective date and rationale. If the analysis reveals a possible legal, safety, fair-housing, insurance, accounting, or security issue, route the underlying case through the approved professional channel rather than attempting to resolve it in the research file.

Interpretation Framework

Interpret findings in layers. First report coverage: how many eligible records were available and complete. Then report process states and durations. Next show exception reasons and reviewer agreement. Finally identify narrow operational questions that management can test. This sequence keeps the quality of the evidence visible beside the result.

A useful conclusion names what the records support, what they do not support, and what evidence would change the interpretation. For example, a concentration of waiting time at one stage may justify a targeted workflow pilot, but it does not establish that one team or outside party caused the delay.

Limitations

Observed timing does not establish legal liability; estimated readings, municipal billing, resident privacy, and provider corrections can alter the record. The proposed review is observational and process-specific. It cannot establish causal effects, legal compliance, financial accuracy, or service quality beyond the evidence and definitions examined.

There is also a survivorship risk when only records still available in the current system are sampled. System migrations, deleted attachments, changed permissions, and manual work outside the platform may reshape the apparent history. Document these boundaries and avoid generalizing beyond the sampled portfolio and period.

Portfolio Application

Begin with a bounded pilot and a prewritten analysis plan. Review the initial results with operations, the data custodian, and any qualified specialist whose authority is implicated. Choose one change, such as a clearer state definition or required source reference, and observe whether evidence completeness improves without creating unmanageable work.

Re-run the same measures after a stable interval. Preserve the original definitions or clearly bridge any changes. If a control creates delay without improving interpretability, revise it. If a simple source link reduces rework, formalize it. Research is valuable when it creates a more testable operating decision, not when it produces a longer report.

Conclusion

Vacancy Utility Charge Leakage Study is best approached as an evidence-design problem. A defensible study defines the unit, separates stages, treats missingness honestly, preserves decision authority, and reports limitations beside results. That discipline gives PortfolioRental teams a practical basis for improving coordination while avoiding claims the data cannot support.

Sources

1. HUD Rental Assistance and Tenant Rights, accessed 2026-10-13.
2. Consumer Financial Protection Bureau Renting Resources, accessed 2026-10-13.
3. NIST Cybersecurity Framework, accessed 2026-10-13.
4. NIST Privacy Framework, accessed 2026-10-13.
5. National Archives Records Management, accessed 2026-10-13.
6. GAO Assessing Data Reliability, accessed 2026-10-13.
7. CISA Secure Our World, accessed 2026-10-13.
8. Department of Energy Energy Saver, accessed 2026-10-13.
9. EPA WaterSense, accessed 2026-10-13.
10. Federal Reserve Small Business Credit Survey, accessed 2026-10-13.