Methodology · Updated 27 May 2026
How we value UAE used cars
Every CarWorth range is produced by the same five-step engine. Below are the exact constants, the data sources, and the edge cases — so you can recreate any number we publish or push back when you think one's wrong.
Inputs
A valuation takes five inputs:
- Make: Brand-level grouping. Variants like "Mercedes Benz" and "Mercedes-Benz" canonicalize to one row.
- Model: Family, not trim. A Camry SE and Camry LE share a model-level median; trim premiums are absorbed into the band width.
- Year: Model year, not registration year.
- Mileage: Bucketed in 30,000 km bands. Linear interpolation inside a band.
- Condition: Self-reported: excellent, good, fair, poor.
Constants
- Asking → sold
- 0.93×
- Condition multipliers
- 1.08 / 1.00 / 0.90 / 0.75
- Depreciation
- 10%/yr
- Mileage bucket
- 30,000 km
- Confidence band
- ±6% to ±18%
- Sample sources
- DubiCars + YallaMotor
Median UAE private close vs list
Excellent / good / fair / poor
Geometric, used for inter-year extrapolation
Linear interpolation inside a bucket
Narrow on dense data; wider on fallbacks
UAE-wide live listings
These live in lib/engine/adapters/car.ts. When we re-tune them (we've done it twice since launch) this page and the engine ship together — there is no version where the prose drifts from the code.
The five-step computation
- Pull every active UAE listing for (make, model, year, mileage bucket) from our sample table. Sources: DubiCars and YallaMotor.
- Take the median of asking prices. Median, not mean — one optimistic seller does not move the answer.
- Multiply by the asking-to-sold ratio of 0.93. UAE private sales close roughly 7% below list after negotiation.
- Apply the condition multiplier: 1.08 excellent, 1.00 good, 0.90 fair, 0.75poor. Symmetric around "good".
- Linearly interpolate within the mileage bucket for the exact km entered, then wrap a confidence band: ±6% when the sample is dense, widening to ±18% on sparse fallbacks.
Data freshness
Sample density per segment is shown on every /car page (look for the "based on N comparable listings" line). The busiest segments refresh daily. The long tail is demand-driven: when a user values a thin-coverage model, that (make, model) gets queued for a fresh pull.
Known limits
- Newest model years: For 2026 model-years still in launch volumes, listing depth is thin. We widen the band to ±12% and label the result "indicative".
- High-spec / enthusiast trims: AMG Black Series, GT3, Trackhawk, etc. trade on collector dynamics the median can't capture. The published band is a floor, not the truth.
- Accident history: Listings rarely disclose category claims. A documented accident historically takes 15-25% off the low for a realistic private-sale offer.
- Emirates spread: Dubai listings skew about 3-5% higher than Sharjah / Northern Emirates for the same car. CarWorth publishes a UAE-wide median, which is appropriate for private sale.
What we do not do
- No buy / sell incentive: We are not a marketplace, not a dealer, not a broker. We earn nothing on the transaction. That's exactly what lets us publish the honest range.
- No paid placements: No make, model, or dealer ever pays for visibility on our /insights or /car pages.
- No single-number quotes: A used-car price is a band, not a number. Anyone quoting one number is either guessing or selling.
Audit and corrections
If a published range disagrees with what you're seeing on the market, email hello@carworth.ae with the make, model, year, and three live listing URLs. We investigate, retune the relevant segment, and ship a corrected range — typically within 48 hours.