# Findings

A job's `findings` are the processor's judgements, one per kind — and, for a fraud assessment, one
per file. Each is the latest version:

| Field | Holds |
| --- | --- |
| `kind` | `missing_information`, `fraud_assessment`, `lender_recommendations` or `deal_insights`. |
| `fileId` | The file a `fraud_assessment` is about; `""` for the others. |
| `version` | Goes up when the processor reports a different answer for the same kind and file. |
| `payload` | The processor's answer, kept exactly as it sent it — below. |
| `at` | When it arrived. |

`findings` is `[]` until the processor has reported. A `payload` may gain keys over time; ignore
the ones you don't use. A `deal_insights` finding is accepted but not sent by the processor today.

## `missing_information`

What the deal still lacks. Present on a finished job; it is why a job ends `needs_information`.

```json
{"kind": "missing_information", "fileId": "", "version": 1,
 "payload": {"completion_status": "incomplete", "completeness_score": 60,
             "missing_documents": ["3 most recent months of business bank statements"],
             "missing_fields": ["credit score"],
             "missing_items": ["3 most recent months of business bank statements", "credit score"],
             "required_actions": "…", "reason": "…",
             "profile_key": "mca_default", "analysis_date": "2026-09-18T14:04:39Z", "…": "…"}}
```

| Key | Holds |
| --- | --- |
| `missing_items` | Everything missing, documents and fields together — what to ask the sender for. |
| `missing_documents`, `missing_fields` | The same, split. |
| `completeness_score` | 0–100. |
| `completion_status` | The processor's one-word verdict. |
| `required_actions` | What to do next, as text. |
| `reason` | Why, in a sentence. |
| `profile_key` | The rule set it was judged by — picked by the `productType` you started the job with. |
| `term_loan_readiness` | For term loans, the readiness assessment; otherwise `null`. |

## `fraud_assessment`

One per PDF bank statement, naming the file in `fileId`.

```json
{"kind": "fraud_assessment", "fileId": "9f0e6a2b-…", "version": 1,
 "payload": {"fraud_score": "35/100 - MEDIUM", "fraud_score_numeric": 35,
             "fraud_reasoning": "…", "bank_name": "JPMorgan Chase"}}
```

| Key | Holds |
| --- | --- |
| `fraud_score_numeric` | 0–100, higher is worse. |
| `fraud_score` | The same, with its level — `LOW`, `MEDIUM`, `HIGH` or `CRITICAL` — for a person. |
| `fraud_reasoning` | Why, as text. |
| `bank_name` | The bank the statement was judged against. |

## `lender_recommendations`

The lenders the processor matched the deal to, when it did.

```json
{"kind": "lender_recommendations", "fileId": "", "version": 1,
 "payload": {"source": "…", "run_id": "…",
             "recommendations": [{"lender_name": "Fitting Funding", "rank": 1, "score": 0.71,
                                  "reason": "…", "requested_amount": 45000,
                                  "requested_term_days": 270, "requested_factor_rate": 1.3}]}}
```

| Key | Holds |
| --- | --- |
| `recommendations` | Best first: `lender_name`, `rank`, `score`, `reason`, `requested_amount`, `requested_term_days`, `requested_factor_rate`, `tags`. A key the processor had no value for is left out. |
| `run_id` | The matching run it came from. |

Nothing is sent to any lender.
