How CarBench scores cars

Methodology and dependability audit · October 11, 2026

Fuel economy and battery range

Battery-electric fuel scores combine the CR/EPA MPG-equivalent efficiency score and published range at 50% each. Range is scored linearly from 0 at 100 miles to 10 at 500 miles, capped at 0–10. EPA range is preferred; manufacturer estimates are labeled where EPA range is not available. Range is full-charge rated range, not a promise of highway or winter distance. Conventional gasoline, regular hybrid, and plug-in hybrid vehicles use CR/EPA MPG because the dataset does not have verified, exact-match full-tank range and tank-capacity inputs; those are not guessed. For battery EVs without CR or EPA efficiency data, a documented independent measured-energy result may be used and is identified as a third-party equivalent, not an EPA rating. A battery EV needs both range and an efficiency input to receive a fuel score; range alone does not generate one.

Dependability

Use an equal average of the three nearest published CR Overall Reliability model-year scores within the same generation. If only two usable numeric historical scores are available, average those two. Earlier years break equal-distance ties; at a generation’s end, use preceding years. Current historical coverage ends at 2025. Keep separate CR hybrid model series separate, and never average across a redesign or a major incompatible powertrain change.

* Brand fallback: When fewer than two usable historical model-year scores are available, use CR’s published brand Predicted Reliability average. This includes all-new generations and models with insufficient numeric coverage. It is a brand estimate, not CR’s rating for that vehicle. † Mitsubishi exception: CR publishes no Mitsubishi brand average in the linked table. Use the exact model-year numeric Overall or Predicted Reliability rating when published, preserving its label. If none is published, show N/A†.

The mean is a CarBench calculation, not a rating published by CR. Raw CR values remain separate. CR’s Overall Reliability, Predicted Reliability, overall vehicle score, road-test score, and owner satisfaction are different measures. Only explicitly displayed reliability ratings out of 100 enter this calculation. A reliability rating out of five is never converted to a made-up numeric reliability rating.

New score = baseline ÷ 10
Used score = max(0, baseline ÷ 10 − min(1, mileage ÷ estimated model lifespan))

The used-car deduction is capped at one CarBench point. Using 40% of estimated mileage life subtracts 0.40 points. This is a small wear allowance, not a failure probability. Model lifespans and reference mileages are estimates. No separate age or warranty deduction is added. If mileage or lifespan is missing, no deduction is fabricated; the audit flags the gap. A used score never exceeds its averaged or fallback baseline; the previous single-year score does not cap the new average.

Other categories

Category scales run from 0 to 10. Defaults total 100%. Compute weighted overall scores at full precision; round only the displayed values. Missing categories are omitted and available weights are normalized. Cost and Safety coverage can reduce their effective weight. Estimates count as available inputs, not verified observations.

Cost

Cost uses the purchase-total anchors described in Selected overrides and purchase cost. Historical depreciation, upkeep and insurance inputs remain available in the source dataset but no longer contribute to Cost. Finance-conditional advertised prices without an unconditional price remain unavailable.

Interior

For CarBench model-years with complete Consumer Reports data, Interior = 25% Front Access, 50% Interior Fit & Finish, and 25% Usability. Each raw CR 1–5 rating is multiplied by 2 only for this derived 0–10 CarBench score. Raw CR values and labels remain in the source dataset. If a required rating is missing, the existing CarBench estimate stays in place. CR's tested trim is named in each score tooltip; exact Canadian trim equivalence has not been verified.

Ride comfort

The 2026 ES 500e cards use Edmunds’ electric ES Comfort rating of 8.7/10 as an explicitly marked electric-family proxy. Edmunds tested the ES 350e, so drivetrain, weight, trim and wheel differences remain.

When all three CR ratings are available, Ride comfort = 50% CR Ride, 20% CR Noise, and 30% CR Front Seat Comfort. Each raw CR 1–5 rating is multiplied by 2 only for this derived 0–10 score. When CR has no complete composite, an explicitly marked Edmunds Comfort rating may be used as a proxy for the 2026 preview/in-test cards where a close model-family review exists. Edmunds scores are retained on their published 0–10 scale and are not presented as CR ratings. The source link beside § identifies the review; trim and wheel-package differences are noted. Otherwise the existing estimate is retained or the category remains unavailable. Edmunds Escalade IQ · Edmunds Lexus ES · Edmunds Navigator.

Space (recalibrated 23 October 2026)

40% cargo capacity + 40% passenger room + 20% loading access. The cargo component is linear between these absolute anchors: 0 L=0, 150 L=2, 300 L=5, 500 L=7, 800 L=9 and 1,100 L or more=10. Thus a 10 cargo score requires roughly 38.8 cubic feet in the scored configuration; a total Space score of 10 also requires top passenger-room and access scores. Scores are capped, not ranked relative to whichever cars are visible.

Two-row vehicles retain seats-up trunk/cargo figures. Three-row SUVs use verified cargo behind the second row with the third row folded; front and second-row seats remain. This is a practical carrying configuration, not a claim that maximum passenger occupancy and that cargo capacity are available simultaneously. Original all-seats-up volumes remain in the source attributes. Minivans use the same configuration where verified; Carnival retains its conservative all-seats-up minimum, which already exceeds the cargo ceiling. Pickups use bed volume. Manufacturer SAE, VDA and publisher measurement methods can differ, especially among older reference values.

Passenger room remains an explicit CarBench class estimate, not a CR score. Full-size Escalade/Escalade IQ/Navigator and minivans score 10 for room; normal three-row families retain their 9.4 class. The G-Class moves to roomy five-seat (9), with 7.5 loading access for its side-hinged gate. Its published two-row cargo volume is retained. The LX Executive VIP is corrected to a four-seat layout rather than being treated as three-row. Loading access is also an editorial layout estimate: large powered-tailgate benchmark SUVs and sliding-door minivans 10, ordinary three-row crossovers 9, G-Class 7.5; other established body-layout estimates remain.

Escalade, Escalade IQ and Navigator reach 10 because all three components reach the calibrated ceiling. This is a useful-capacity benchmark, not a claim of perfect ergonomics. The same formula applies to every card. Complete per-vehicle before/after audit and source links.

Steering

Scores how appropriately weighted, direct, accurate and communicative the steering feels. 0 = very heavy and numb; 5 = ordinary mainstream steering; 10 = light, sharp, accurate and rich in feedback. Scores are model-family estimates where comparable road tests are unavailable; wheel and trim differences can affect feel. The Mazda3 is scored highly for precise, well-weighted steering; the Corolla is more modest for light steering with limited feel; the Civic scores well for quick, well-weighted steering. Car and Driver 2026 Mazda3 · APA 2026 Corolla · Consumer Reports 2025 Civic · Consumer Reports 2024 Model Y.

Acceleration

Acceleration is 10 at 3 seconds 0–60 mph, 5 at 8 seconds, and 0 at 13 seconds or slower, with a straight-line scale between. CR measured times take priority; when unavailable, an independent road-test time may be used if the model year and powertrain match. A clearly marked same-generation prior-year test or manufacturer-stated figure is used only where noted. The § link beside a score opens its source, and the calculation remains 13 minus published 0–60 seconds, clamped to 0–10. Results vary with trim, tires, rollout and test procedure.

Fuel economy

Fuel economy uses Consumer Reports CR MPG, Owner Reported MPG, or Ratings & Specs Overall Mileage when available, with EPA combined MPG/MPGe as fallback. For four 2027 entries without matching published figures, a documented 2026 configuration supplies a CarBench planning estimate. If neither a source nor an appropriate documented reference is available, the card shows N/A and excludes that category from its weighted overall score. For an electric vehicle with no CR or EPA efficiency figure (2026 Cadillac Escalade IQ, which is above the EPA light-duty weight limit), Edmunds’ measured real-world energy use is converted to an MPGe equivalent (33.705 kWh per gallon) and marked as a third-party measurement, not an EPA rating. The straight-line score is 0 at 15 L/100 km (or Le/100 km) and 10 at 2 L/100 km (or Le/100 km), capped to 0–10. PHEVs use charged-operation energy-equivalent efficiency for the score and show gasoline-only consumption separately. Stored combined US mpg is converted with 235.214 ÷ mpg. Acceleration times for some trims are estimates; verify a trim-specific road test before comparing close scores. The listed acceleration inputs were checked against published road tests where a close powertrain match exists; results can still vary by trim, tires, rollout and test procedure.

Grip (snow/mud)

Grip is a CarBench capability estimate: 70% winter-road traction and 30% mud/rough-ground capability. It considers the fitted AWD/4WD system, traction control, clearance, low-range gearing and confirmed locking/torque-vectoring hardware. Snow comparisons assume comparable suitable winter tires; AWD does not shorten tire-limited stopping distances. Scores distinguish exact trims and do not award optional off-road equipment. CarBench ceilings are 6 for FWD, 4 for RWD, and 9 for AWD/4WD without verified exceptional hardware or system performance. A score above 9 requires documented exact-trim capability, such as effective front/rear locking differentials or an equivalently capable system; nameplates alone do not qualify. These ceilings apply to this snow/mud capability estimate, not dry-road handling or braking.

Handling and Braking

Handling = 2 × (60% CR Routine Handling + 40% CR Emergency Handling). Braking = 2 × CR Braking. Preserve the raw 1–5 ratings. When complete components are unavailable, retain clearly identified model-family estimates. Tire choice, trim, and test conditions can affect comparisons. When CR has not tested a model but has tested its same-body hybrid twin in the same model year (2026 Lincoln Nautilus), the twin’s CR ratings are used and labelled as a proxy. Third-party Handling (added October 21, 2026): when no CR rating exists, a published instrumented skidpad result (Edmunds, Car and Driver or MotorTrend) is converted with a calibration fitted on 17 CR-tested vehicles: Handling = −10.08 + 20.68 × skidpad g, capped 0–10 (r = 0.87; typical error about ±0.9 points). These scores are estimates, and each one names its test and source. Published braking distances were checked the same way and rejected: across 21 CR-tested vehicles they predicted CR Braking only weakly (typical error 1.7 points against a 2.0-point spread), so those distances are not used to reconstruct CR Braking. The user-approved independent dry-stop scale below is a separate CarBench estimate. Historical Edmunds pages sometimes publish native Interior subratings, which take priority over editorial judgments when the cabin matches.

Safety

Uses the crash-test sources attached to each vehicle, including IIHS test domains and NHTSA overall stars. An applicable NHTSA overall five-star rating maps to 10/10. IIHS evidence uses the available tested domains and coverage adjustment. Different testing protocols are not interchangeable, and a maximum score does not imply identical crash protection. Unavailable applicable crash results remain missing. Exact grades and sources are retained with each card.

Headlights

2027 headlight carryovers marked § are estimates using the 2026 model-family result and current equipment information. Source year and lamp-group assumptions are retained; these are not new IIHS or CR tests of the 2027 vehicle.

Headlights carry 4% of the current default weight. Current defaults are Cost 16%, Dependability 18%, Safety 6%, Headlights 4%, Fuel economy 8%, Space 8%, Interior 8%, Ride comfort 8%, Grip 8%, Handling 4%, Braking 4%, Acceleration 4%, and Steering 4% (100% total). Consumer Reports is used first: the displayed 1–5 rating is retained in the source data, then multiplied by two for CarBench’s 0–10 comparison score. Where CR has no numeric rating, an applicable IIHS headlight grade is used with CarBench’s existing mapping: Good 10, Acceptable 7, Marginal 3, Poor 0. For IIHS pages with several trim or headlamp variants, CarBench uses the matching grade when trim equivalence is clear; otherwise it uses the lowest applicable grade. U.S.-market equipment may differ from Canadian equipment. Ratings outside IIHS stated model-year applicability and vehicles with no numeric source are shown as N/A and omitted from the overall score. Card-level raw ratings, sources and grade mappings are retained in the project audit data.

Every vehicle’s dependability calculation

301 CarBench vehicles. See the versioned dependability audit for method counts; model-year numeric CR predictions take priority over brand fallback on the 2026–2027 luxury cards. Baselines below use the 0–10 CarBench scale; source ratings retain CR’s 0–100 scale. Two-year means reflect available numeric coverage in the audited generation window.

New hybrid SUVs and minivans

Added October 11, 2026. Current Canadian model benchmarks use manufacturer or dealer prices before Ontario HST and the existing $59 licensing allowance. Stock availability and prices can change. CR scores use the displayed U.S. tested trim for the same model year and powertrain; exact Canadian trim equivalence is unverified. Source ratings retain their original scales. Grip and steering are CarBench planning estimates.

Three-row cargo uses published volume behind the third row with all passenger seats upright. Minivans use a 9/10 loading-access estimate and the existing 9.4/10 three-row passenger-fit class. CAA exact model-year depreciation and upkeep are unavailable for these additions, so the existing Cost coverage rules omit those inputs. The Sorento uses the earlier applicable 2026 moderate-overlap result conservatively until its manufacture date is verified; the improved Good result applies to builds after September 2025.

Additional electric vehicles (October 11, 2026)

CR fuel observations retain the exact “Overall Mileage” MPGe label. They represent energy-equivalent efficiency, not gasoline MPG or electricity cost. CR tested U.S. trims are model and powertrain proxies. Canadian Kona Electric uses CR’s 2025 test because no 2026 U.S. page exists. Its electric-specific crash rating was unavailable, so Safety is omitted. Gasoline Kona crash results are not substituted.

New EV dependability uses historical Overall Reliability averages where at least two same-generation scores exist. Otherwise, the published CR brand fallback is marked *. Current-year Predicted Reliability remains separately recorded. No EV incentive, conditional order rebate, or expired discount is deducted. Manufacturer/dealer prices are reference benchmarks; orderable models are not claims of immediate dealer stock. Grip and steering remain planning estimates. Missing exact CAA cost components are omitted.

Sources and limits

CR model-year ratings and road tests can use U.S. trims that differ from Canadian listings. Exact engine, transmission, drivetrain, equipment, and tire equivalence is not verified for every entry. Separate CR hybrid model series are retained. Manufacturer references supplement CR history for recent redesigns and major powertrain changes. Model-year summaries can combine multiple trims and engines.

Purchase totals estimate advertised price plus 13% Ontario HST and $59 licensing. Dealer fees and cash/finance conditions may differ; inventory and prices can change. CAA annual values are model-year family estimates, not odometer-specific predictions. Missing cost components are omitted. Fuel is scored separately from Cost to avoid weighting the same expense twice.