We combine listing signals, EV-specific risk checks, market context, and routine-fit analysis to help you decide whether a car is worth pursuing.
Three surfaces, each with its own analysis logic.
A physics-informed lookup with empirical priors — not a trained ML model.
We don't pull a real-time SOH value from the car, and we don't have a trained predictive model with held-out test sets. What we build is an expected health range using a small set of raw inputs combined with published degradation priors for each battery chemistry.
Raw inputs — what actually goes in
What we cannot track from a listing
These inputs move individual packs meaningfully. We don't have them. The range we output is the population distribution for that VIN profile — not a prediction for that specific unit.
Degradation priors by chemistry
NMC / NCM
~2.0% / year
Majority of current EVs — Hyundai, Kia, VW, Ford
LFP
~1.5% / year
Tesla Standard Range, BYD — better longevity tail
NCA
~2.3% / year
Older Tesla Model S/X pre-2021
Air-cooled
~3.0% / year
Pre-2023 Nissan Leaf — steeper decline cliff
Rates sourced from published fleet studies and peer-reviewed degradation literature. Chemistry differentiation matters most at the tails (LFP longevity advantage, air-cooled Leaf degradation cliff) and less for mainstream NMC-to-NMC comparisons.
What the output actually means
The result is a cohort range, not a unit measurement. “88–93% expected at this mileage” means: vehicles with this VIN profile at this mileage typically land in that band based on population-level degradation data.
The signal is most useful for relative comparison — two otherwise identical listings where one is at 90k miles and one at 45k miles, or where one is a Leaf (air-cooled) and one is a Bolt (liquid-cooled). It's not a warranty-grade SOH certificate.
On OBD2 SOH readings: They're noisy — a February reading in a cold climate is not the same as a post-summer full-cycle reading. Our estimate doesn't use OBD data. It's a population prior applied to VIN-decoded specs, which is why it doesn't vary with season or charge state.
OFFO looks at three things — not just specs.
Vehicle condition and risk — Is there evidence this car will cause problems?
Deal economics — Is the price fair given what the listing tells us?
Ownership fit — Does this car match the buyer's real routine?
A car can be mechanically sound but still a bad deal.
A fair deal can still be the wrong fit for someone's routine.
Auction vehicles may have parts value even when repair is not recommended.
What goes into each analysis.
Vehicle & safety
Market & pricing
EV-specific ownership
User-provided
Some data is exact. Some is inferred from known patterns. Confidence reflects that.
What happens after you paste a URL or answer a question.
We extract key details from the listing, compare price against similar vehicles in context, and check for EV-specific warning signs — missing battery proof, weak service history, unclear charging details, and ownership-risk patterns. We generate a verdict and seller questions based on what is missing or risky.
We use signals, checks, and weighted inputs — not exact condition data.
Confidence matters as much as the score.
What can improve confidence
Verified battery health · Service records · Title history · Exact trim · Charging setup details · Climate & parking details · Inspection results