Methodology
How we tell a fair price from a bad one.
Buying a used car in the UAE is overwhelming — same model, same year, same mileage, and prices that vary by 40%. UAE Motors solves one problem: telling you whether the listing in front of you is below, at, or above what the market actually pays.
The model in one paragraph
Where the data comes from
Largest UAE listings inventory. Behind Imperva/Incapsula; we route via Firecrawl when scraping.
Dealer-vetted listings with structured JSON-LD (images, body type, colour). Smaller volume but higher data quality.
Saudi-focused. Many sellers omit prices entirely; we only ingest listings with a parseable SAR amount.
We respect every source's rate limits and Terms of Service. We do not republish listing photos or contact details — we link back to the source.
Frequently asked questions
Are you a marketplace?
No. We don't host listings, take commissions, or contact sellers on your behalf. We aggregate listings from public marketplaces (Dubizzle, Carswitch, haraj.com.sa) and overlay statistical analysis. Every car we display links back to its source where you transact directly.
How is the deal score computed?
For every listing we identify a peer segment — same make, model, year, and mileage band — within the same region. We compute the segment's median price and mean absolute deviation (MAD), then score the listing as a modified z-score: 0.6745 × (price − median) ÷ MAD. We clip extreme values at ±5 to keep the UI sane on thin segments.
What does Excellent / Good / Fair / Above market / Overpriced mean?
Excellent: z < −2 (≈21% under median). Good: z < −1. Fair: |z| ≤ 1. Above market: z ≤ 2. Overpriced: z > 2. Suspect is a special bucket for prices below 50% of median or above z > 3.5 — almost always data errors or accident-damaged cars.
What is 'confidence'?
It tells you how many comparable listings the score is based on. High = 10+ comparables. Medium = 5–9. Low = 4. None = under 4 (we then refuse to assign a strong label). We never surface excellent_deal labels with confidence below medium.
What buyer warnings do you issue?
Mileage out of band for the model year (e.g. 200k km on a 2023 car or 5k km on a 2014 car), known model-specific issues from a curated knowledge base (transmission shudders, oil consumption, recall windows), and listings flagged 'suspect' by our pricing model. We never invent warnings — every entry is tied to a documented source.
How fresh is the data?
Daily scrapes at 02:00 UTC (Dubizzle), 03:00 UTC (haraj.com.sa for KSA), and 04:30 UTC (Carswitch). Each listing has a `last_seen` timestamp; lifecycle tracking marks listings as 'gone' when the source returns a 404.
Why don't you have photos for every listing?
Carswitch listings include structured image URLs in their JSON-LD; we surface those. Dubizzle and haraj listings don't expose images in a stable format yet, so we link to the source page where you can browse the gallery directly.
Do you store my searches or personal data?
We don't have user accounts yet. Filter parameters live in URL query strings only — no cookies, no analytics beyond aggregate page-view counts, no email captures. When we add saved-search alerts, opt-in will be explicit.
Ready to find a fair deal?