Cross-Org Deduction Pattern Intelligence Report
Cross-Org Deduction Pattern Intelligence Report
Overview
The Cross-Org Deduction Pattern Intelligence Report is a monthly automated workflow (introduced in v0.1.383) that analyses dispute outcomes and deduction behaviour across the entire DepositClear platform. It gives the admin team a high-level, data-driven view of how disputes are being decided, which deduction types carry the most financial risk, and where AI-assisted recommendations are performing well or need refinement.
Schedule
The report runs automatically via cron on the 3rd of every month at 03:00 UTC.
cron: 0 3 3 * *
No manual triggering is required. The workflow is fully automated.
What the Report Contains
1. Top 10 Adjudicator Reasoning Tags
Aggregates all resolved dispute outcomes platform-wide and surfaces the 10 most frequently cited reasoning tags used by adjudicators when making their decisions. This helps identify which evidence types, arguments, and deduction justifications consistently hold up — or fail — under scrutiny.
2. Highest-Loss Deduction Categories
Identifies the deduction categories (e.g. cleaning, damage, redecoration, rent arrears) that result in the greatest financial loss to landlords or agents when disputes are raised and decided against them.
3. Average Award Ratios by Category and Property Type
Breaks down the average proportion of a claimed deduction that is actually awarded, segmented by:
- Deduction category (e.g. cleaning, wear & tear, damage)
- Property type (e.g. flat, house, HMO)
This helps identify where claims are being systematically over- or under-valued relative to what adjudicators award.
Data Sources
| Entity | Role |
|---|---|
dispute_outcomes | Source of adjudicator decisions and reasoning tags |
deposit_release_deductions | Deduction records used for category and amount analysis |
organizations | Scopes data across all orgs for platform-level aggregation |
ai_performance_snapshots | Stores the computed report results |
users | Identifies admin recipients for the email report |
Output & Delivery
Platform Snapshot
Computed metrics are written as an updated record in ai_performance_snapshots at the platform level. This record can be used to track AI recommendation performance over time and to inform model improvements.
Admin Email Report
Upon completion, the workflow automatically emails all users with role = 'admin' on the DepositClear platform. The email contains the full report summary including all three metric sets described above.
Who Receives This Report
Only users with the admin role receive the monthly email. This is scoped to the DepositClear internal admin team — not individual organisation admins.
Use Cases
- Platform health monitoring: Track whether dispute outcomes are improving or degrading over time.
- AI model calibration: Identify deduction categories or property types where AI award predictions deviate significantly from actual adjudicator decisions.
- Risk guidance: Inform product decisions around which deduction categories to flag as high-risk within the landlord/agent workflows.
- Compliance oversight: Maintain a longitudinal record of dispute patterns relevant to Renters' Rights Act compliance monitoring.