Research question and scope

Published September 10, 2026.

This study asks whether courtesy credit decisions are consistent with the organization's declared recovery rules after accounting for documented customer impact, issue type, prior recovery, and agent authority. It covers discretionary credits in named queues during a fixed period. Required refunds, charge corrections, and statutory remedies are excluded.

Methodology

Publish the eligibility rules and analysis plan before reviewing outcomes. Extract a reproducible sample stratified by issue family, channel, and decision. Define variables from policy rather than selecting them after seeing results. Reviewers should code verified impact, policy eligibility, amount, approval level, repeat credit history, and documented rationale.

Use a blinded second review for a subset of cases, report agreement, and resolve ambiguous categories under a written adjudication rule. Protect customer and agent identities in the analysis file. Analyze missing documentation as its own result.

Measures and analysis

Report approval rate, credit value relative to documented impact, policy exceptions, and rationale completeness. Present distributions rather than averages alone. Use adjusted comparisons only when the model, reference groups, missing-data treatment, and uncertainty are disclosed. Examine operational groups for process improvement, not individual performance ranking.

Inference boundaries and limitations

Observed differences may reflect unrecorded facts, customer preference, product rules, or supervisor judgment. The study cannot infer discrimination, intent, customer satisfaction, or causal impact on retention from case records alone. Results are limited to included queues and the policy in force during the study period. Small groups may require suppression for privacy and statistical stability.

References

  1. US Government Accountability Office, Designing Evaluations
  2. US Government Accountability Office, Assessing Data Reliability
  3. National Institute of Standards and Technology, AI Risk Management Framework
  4. Federal Trade Commission, Consumer Reviews and Testimonials Rule