> ## Documentation Index
> Fetch the complete documentation index at: https://docs.majarrah.io/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Decisions

> How Majarrah's decision engine works and what decisions cost.

## What is a decision?

A decision is a structured verdict on a property. Instead of just showing you a listing, Majarrah reasons about it — comparing price against market benchmarks, scoring the location, evaluating the specs, and returning a clear verdict.

Every decision returns:

* A **verdict** - `match`, `partial`, or `no_match`
* A **score** from 0 to 100
* A **breakdown** across price, location, size, and property-specific details
* Optionally, a **reasoning** paragraph in Arabic or English

## Two modes

<CardGroup cols={2}>
  <Card title="Scoring" icon="chart-bar">
    Pure algorithmic comparison against market data. Returns the score and breakdown instantly. Costs **1 decision**.
  </Card>

  <Card title="Reasoning" icon="brain">
    Scoring + AI explanation. Adds a plain-language paragraph explaining the verdict. Costs **3 decisions**.
  </Card>
</CardGroup>

## Decision packs

Decisions are purchased in packs. All packs include a **24-hour refund guarantee**.

| Pack     | Decisions | Best for               |
| -------- | --------- | ---------------------- |
| Starter  | 5         | Evaluating a shortlist |
| Standard | 20        | Active property search |
| Pro      | 50        | Brokers and investors  |

You can buy packs from your [billing page](/brokers/billing) inside the dashboard.

## How scoring works

The engine scores every property across four weighted buckets:

| Bucket   | Weight | What it measures                                                   |
| -------- | ------ | ------------------------------------------------------------------ |
| Price    | 30%    | Property price vs. market average for the location, type, and size |
| Location | 20%    | Area demand, market trend, proximity to preferences                |
| Size     | 20%    | Price per sqm vs. market benchmark                                 |
| Details  | 30%    | Type-specific factors: bedrooms, floor, elevator, zoning, etc.     |

Each bucket returns a score from 0–100 and a status (`pass`, `warn`, or `fail`). The weighted average becomes the final score.

**Verdict thresholds:**

* 70–100 → `match`
* 40–69 → `partial`
* 0–39 → `no_match`

## Confidence

The confidence level (`high`, `medium`, or `low`) reflects how much data was available to make the decision. Providing more property details raises confidence.

<Note>
  Decisions never go stale — if you want a fresh verdict on the same property, simply request a new one. Market benchmarks update continuously.
</Note>