What was missing when the decision was made?
Application and feature events known as of 2024-10-01. Every expected feature slot remains in the denominator, including unavailable inputs. No outcome file or fitted model is used.
Coverage by input
Available means observed and available by the application's decision. Late means a pre-decision observation first arrived after that decision but by the report cutoff. Post-decision means only later observations are visible. No visible event means nothing was available by the report cutoff, even if the supplied file contains future records.
Swipe or scroll the tables horizontally to inspect every column. Counts use all applications in each configured population, without dropping incomplete cases.
| Population | Input | Available / expected | Late | Post-decision | No visible event |
|---|---|---|---|---|---|
| reference | utilization | 480/480 | 0 | 0 | 0 |
| reference | debt_to_income | 480/480 | 0 | 0 | 0 |
| reference | prior_delinquencies | 480/480 | 0 | 0 | 0 |
| current | utilization | 318/320 | 0 | 0 | 2 |
| current | debt_to_income | 319/320 | 1 | 0 | 0 |
| current | prior_delinquencies | 319/320 | 0 | 1 | 0 |
Current origination months
These are observed counts, including small months. An empty population has no coverage rate; it is not treated as 0% missing or 100% complete.
| Vintage | Input | Available / expected | Late | Post-decision | No visible event |
|---|---|---|---|---|---|
| 2024-07 | utilization | 78/80 | 0 | 0 | 2 |
| 2024-07 | debt_to_income | 79/80 | 1 | 0 | 0 |
| 2024-07 | prior_delinquencies | 79/80 | 0 | 1 | 0 |
| 2024-08 | utilization | 80/80 | 0 | 0 | 0 |
| 2024-08 | debt_to_income | 80/80 | 0 | 0 | 0 |
| 2024-08 | prior_delinquencies | 80/80 | 0 | 0 | 0 |
| 2024-09 | utilization | 80/80 | 0 | 0 | 0 |
| 2024-09 | debt_to_income | 80/80 | 0 | 0 | 0 |
| 2024-09 | prior_delinquencies | 80/80 | 0 | 0 | 0 |
| 2024-10 | utilization | 80/80 | 0 | 0 | 0 |
| 2024-10 | debt_to_income | 80/80 | 0 | 0 | 0 |
| 2024-10 | prior_delinquencies | 80/80 | 0 | 0 | 0 |
Visible timing in days
Availability lag is availability date minus observation date, measured on the selected on-time events. Lateness uses the first visible pre-decision observation that arrived late, measured from the decision. A dash means no applicable records; an unknown delay is not filled with zero.
| Input | On-time events | Median lag | Maximum lag | Median lateness | Maximum lateness |
|---|---|---|---|---|---|
| utilization | 318 | 4.0 | 4 | — | — |
| debt_to_income | 319 | 4 | 4 | 3 | 3 |
| prior_delinquencies | 319 | 4 | 4 | — | — |
Inspect unavailable inputs
Showing 4 of 4 unavailable slots across both populations. Download the full application-feature ledger.
| Application | Input | Reason | Decision | Observed | Available |
|---|---|---|---|---|---|
| CV202407-0000 | utilization | No visible event | 2024-07-01 | — | — |
| CV202407-0001 | debt_to_income | Arrived after decision | 2024-07-01 | 2024-06-26 | 2024-07-04 |
| CV202407-0002 | prior_delinquencies | Observed after decision | 2024-07-01 | 2024-07-02 | 2024-07-03 |
| CV202407-0003 | utilization | No visible event | 2024-07-01 | — | — |
This diagnoses decision-time event coverage, not raw CSV null rates, all forms of data quality, credit risk or model deterioration. Malformed values and duplicate revisions still fail validation. Existing scoring and distribution checks still reject missing decision-time features. A successfully created or verified diagnostic report can contain incomplete coverage.