AI-Powered Vehicle Risk Analysis

What You Get — The 4-Factor Analysis

1

This Vehicle

What stands out about this specific car? We analyse its full MOT history — how many tests, how many failures, what defects have appeared, whether they're recurring, and how its mileage tracks over time. This gives us the baseline.

2

Similar Vehicles

We pull data from up to 200 comparable vehicles — same make, model, fuel type, and age bracket. We look at what defects are most common across the group, which systems fail most often (brakes, suspension, emissions, etc.), and what the typical failure rate looks like.

3

How They Compare

We overlay your vehicle's actual defect history onto the group pattern. Does this car have more brake advisories than typical? Fewer emissions issues? A higher failure rate? The comparison is quantified — not "a bit worse" but "1.8× the group average."

4

What It Means For You

We translate the data into actionable insight. If this vehicle has a pattern that correlates with expensive repairs, we say so. If it's tracking below average risk, we say that too. No hedging — just evidence-based assessment using safe, comparative language.

The Technology Behind It

Data collection — We fetch the target vehicle's full MOT and defect history, plus data from up to 200 comparable vehicles. Defect records, failure rates, mileage, and test outcomes are all gathered.
Defect pattern analysis — We query pre-aggregated defect pattern tables that tell us the top 30 most common defects for this vehicle group, their frequencies, and severity distribution (advisory vs. minor vs. major vs. dangerous).
Semantic embedding (MiniLM) — We use a sentence-transformer model (all-MiniLM-L6-v2) to convert every defect description into a mathematical vector. Defects with similar meaning cluster together — "brake pad worn" and "brake pad thickness below minimum" are recognised as the same issue. This lets us detect patterns that simple keyword matching would miss.
Prediction generation — The model identifies the 10 most likely defects for your vehicle based on what similar vehicles experienced. Each prediction comes with a confidence score and a frequency (how common it is across the peer group).
Risk scoring — We calculate a 0–100 risk score from four components: failure rate vs. group average (0–40 points); defect frequency per MOT vs. group (0–30 points); dangerous defect history (0–20 points); mileage trajectory anomalies (0–10 points).
Verdict — The score maps to a clear assessment: "Lower risk than average", "Typical for its class", or "Higher risk than average" — always with evidence explaining why.

Example Output

2018 BMW 320d — AI Hunch Results

Risk Score: 62/100 — Higher risk than average

Top 3 predicted defects:

Exhaust emission lambda reading not within limits (78% confidence — seen in 45% of peers)
Brake disc worn below minimum thickness (71% confidence — seen in 38% of peers)
Suspension arm ball joint dust cover damaged (65% confidence — seen in 31% of peers)

Verdict: This vehicle has a 28% MOT failure rate vs. 16% for the peer group. Emission-related defects appear at 1.7× the group frequency. Brake wear is tracking above average. Budget for emission system and brake work within the next 12 months.

What This Is Not

1

Not a crystal ball

We predict likely issues based on statistical patterns. A high-risk score doesn't mean the car will definitely fail. A low-risk score doesn't mean it's perfect.

2

Not a replacement for an inspection

Always get a physical inspection before buying. Our analysis tells you what to ask the mechanic to look at.

3

Not based on opinions

No forums, no anecdotes, no manufacturer claims. Only structured DVLA data processed through statistical models.

Built on complete MOT history analysis — see how we structure every test and defect before the AI runs.