Reason codes

# Reason codes

What each answer rests on, in `reasons`, in the order the decision adds them.

Built and tested, not yet deployed. Early access.

## Restricted merchant

`restricted_merchant`

| Code | When |
| --- | --- |
| `mcc_<code>` | An MCC was sent, for example `mcc_7995`. |
| `rule_gambling_category` | The MCC is a network gambling code (7995, 7800, 7801, 7802, 7948, 9754). |
| `rule_known_brand` | The name matched our gambling brand list, without a gambling MCC. |
| `kind_sportsbook` | The model’s answer to “what kind of business is this”, when it gave one. |
| `kind_online_casino` | The model’s answer to “what kind of business is this”, when it gave one. |
| `kind_sweepstakes_casino` | The model’s answer to “what kind of business is this”, when it gave one. |
| `kind_lottery` | The model’s answer to “what kind of business is this”, when it gave one. |
| `kind_fantasy_or_prediction_market` | The model’s answer to “what kind of business is this”, when it gave one. |
| `kind_race_wagering` | The model’s answer to “what kind of business is this”, when it gave one. |
| `kind_skin_gambling` | The model’s answer to “what kind of business is this”, when it gave one. |
| `kind_not_gambling` | The model’s answer to “what kind of business is this”, when it gave one. |

## Spend type

`spend_type`

| Code | When |
| --- | --- |
| `mcc_<code>` | Always. |
| `mcc_unlisted` | The code isn’t in our table of 981 descriptions, so the model saw “unknown”. |

## Receipt to charge

`receipt_match`

| Code | When |
| --- | --- |
| `amount_exact` | The chosen charge equals the receipt total (within half a cent). |
| `amount_within_3pct` | Within 3% of the total. |
| `amount_differs` | More than 3% away. |
| `name_match` | Name similarity of 0.7 or more. |
| `name_differs` | Name similarity under 0.7. |
| `within_a_day` | Charged within 24 hours of the receipt. |
| `within_3_days` | Within 72 hours. |
| `over_3_days` | More than 72 hours apart. |
| `no_single_charge` | The answer is `"none"`. |
| `rule_agrees` | The model’s pick equals our engine’s own pick (or both say none). |
| `rule_differs` | It doesn’t. Review these cases first. |
| `rule_no_purchases` | Settled without the model: no positive, non-declined transaction. |

## Refund to purchase

`refund_pairing`

| Code | When |
| --- | --- |
| `full_refund` | The share returned is within 0.5% of the purchase. |
| `partial_refund` | Less than the purchase. |
| `refund_exceeds_purchase` | More than the purchase. |
| `name_match` | Name similarity of 0.7 or more. |
| `name_differs` | Name similarity under 0.7. |
| `within_45_days` | Refunded within 45 days of the purchase. |
| `over_45_days` | More than 45 days after. |
| `no_refund` | The answer is `"none"`. |
| `rule_agrees` | The model’s pick equals our engine’s own pick (or both say none). |
| `rule_differs` | It doesn’t. Review these cases first. |
| `rule_no_refunds_after_purchase` | Settled without the model: no refund after the purchase. |

## Any decision

| Code | When |
| --- | --- |
| `model_declined` | The model declined the question. `answer` and `probability` are `null`; `abstain` is true. |

## Your own question

`reasons` is empty, `[]`, unless the model declined; then it is `["model_declined"]`.

## Review first

> Note. Use `rule_differs` to pick which cases to review first. Today it comes back only from receipt to charge and refund to purchase.

Source: /api/docs/reasons