Auto Insurance Telematics & UBI 2026 — Progressive Snapshot, Allstate Drivewise, Root, Metromile, CMT, Arity, Carrot Permile, SOMPO Drive Save Deep Dive (The Year PHYD/PAYD Became the Standard)

Prologue — "You Had Two Hard Brakes Yesterday at 17:33"

A morning in April 2026. A driver in Chicago opens the Progressive app and a notification pops up: "You logged two hard brakes on Cannon Street yesterday at 17:33. Your weekly score is 84." A passenger asks, "How does the insurance company know that?"

The answer is the Snapshot telematics they enabled six months ago. The smartphone's accelerometer and GPS stream data every second, and Progressive's scoring model classified those samples as longitudinal deceleration above 0.3g (harsh braking). That score is what drives the average 17% and up-to-30% discount at renewal.

This is how 2026 auto insurance works. Premium is decided by how much you drive (PAYD, Pay As You Drive) and how you drive (PHYD, Pay How You Drive). All US top-5 carriers run a UBI line. Korea's Carrot Permile has crossed one million cumulative subscribers. Japan's SOMPO Drive Save just hit its five-year anniversary. Tesla Insurance sets premiums from its own real-time driving score.

This post maps the 2026 telematics + UBI landscape — from Progressive Snapshot to Tesla Insurance, the technical differences between OBD-II, smartphone, and embedded telematics, the scoring algorithms, data privacy, and what Korea and Japan actually look like.

Chapter 1 · The Two Branches of UBI — PAYD and PHYD

First, the vocabulary. Usage-Based Insurance has two main branches.

PAYD (Pay As You Drive) — Look only at distance. A rate per mile or per kilometer. Metromile is the canonical model. Favors people who drive less.
PHYD (Pay How You Drive) — Look at how you drive. Behaviors like harsh braking, speeding, phone use, and cornering are scored and discounted. Progressive Snapshot, Allstate Drivewise, and State Farm Drive Safe & Save sit here.

The de facto 2026 standard is PHYD + mileage correction. Snapshot looks at miles too. Drivewise looks at driving time too. Pure PAYD has lost share; Metromile, after Lemonade's acquisition, was absorbed into a closed channel.

On top of that there is one more axis: continuous monitoring. Progressive monitors only the first six months and then locks the score; Root and Tesla Insurance recompute the score at every renewal. That difference governs how volatile premiums are and how the user experiences the product.

Model	Data	Pricing	Carriers
PAYD	Distance	Per mile / per km	Metromile, Carrot Permile
PHYD	Behavior score	Discount %	Progressive Snapshot, Allstate Drivewise
Hybrid PHYD+PAYD	Score + distance	Score weighted by miles	Root, Tesla Insurance
Continuous PHYD	Re-evaluated per renewal	Annual variation	Root, Tesla Insurance

Chapter 2 · Three Ways to Collect Data — OBD-II, Smartphone, Embedded

Where does UBI data come from? Three approaches coexist in 2026.

OBD-II dongle — A small device plugged into the diagnostic port (mandated in the US since 1996). Snapshot Gen 1 (2008–2017) and early Carrot Permile used this. Accurate, but heavy on the user and on the carrier's device budget.
Smartphone telematics — Read the accelerometer, gyroscope, GPS, and phone-use events through OS APIs. The user just installs an app. About 80% of the US market in 2026 uses this. CMT and Arity supply the SDKs.
Embedded telematics — A modem built into the vehicle from the factory sends data directly. Tesla, BMW ConnectedDrive, GM OnStar, and Hyundai Bluelink follow this model. The most accurate, but requires OEM partnerships.

The smartphone approach has two weaknesses. First, driver attribution is unknown — you don't know who was driving. Second, if the phone isn't in the car, no data is captured. To address this, CMT runs a hybrid with a Bluetooth Low Energy beacon (DriveWell Tag) installed in the car, measuring only when the phone is present.

Approach	Accuracy	User burden	Cost/unit	Carriers
OBD-II	High	Mount device	20–40 USD	Early Metromile, Carrot Gen 1
Smartphone	Med-high	App only	0 USD	Snapshot, Drivewise, Root
Smartphone + BLE beacon	High	One-time beacon	5–10 USD	CMT DriveWell
Embedded	Highest	None	0 USD	Tesla, BMW, OnStar

Chapter 3 · Progressive Snapshot — The Original and the Standard

Progressive filed its first UBI patent in 1998 and launched Snapshot commercially in 2008. By 2026, with over 40 million cumulative enrolled vehicles, Snapshot is the de facto standard of the US UBI market.

Snapshot's pricing model is simple. After enrollment, a six-month monitoring period determines the discount at renewal. 17% average, 30% maximum is the marketing line. But rate increases also happen — about 20% of Snapshot users see a rate hike.

The four core inputs into the score are:

Hard Braking — Longitudinal deceleration above 0.3g (corrected for vehicle weight and speed).
Time of Day — Driving between midnight and 4 AM is penalized due to higher accident rates.
Mileage — Adjusted around a 12,000-mile annual baseline.
Phone Use — Touching the screen while driving (introduced in 2018).

Progressive does not publish the score algorithm, but academic analyses indicate hard braking carries the largest weight. So smooth stopping is the single most important behavior for UBI.

# Simplified hard braking detection — evaluate accelerometer time series in 1s windows
import numpy as np
from dataclasses import dataclass

@dataclass
class AccelSample:
    t: float        # epoch seconds
    ax: float       # longitudinal m/s^2 (positive = forward acceleration)

HARSH_BRAKE_G = 0.3
G = 9.80665

def detect_harsh_braking(samples: list[AccelSample], window_s: float = 1.0) -> list[float]:
    """Return list of event timestamps where peak deceleration >= 0.3g in a 1s window."""
    events = []
    if not samples:
        return events
    samples = sorted(samples, key=lambda s: s.t)
    n = len(samples)
    j = 0
    for i in range(n):
        while j < n and samples[j].t - samples[i].t < window_s:
            j += 1
        window = samples[i:j]
        if not window:
            continue
        # longitudinal deceleration is -ax; peak deceleration in window
        peak_decel_g = max(0.0, -min(s.ax for s in window) / G)
        if peak_decel_g >= HARSH_BRAKE_G:
            events.append(window[0].t)
    # deduplicate consecutive events in same braking arc
    deduped = []
    for t in events:
        if not deduped or t - deduped[-1] > window_s:
            deduped.append(t)
    return deduped

Chapter 4 · Allstate Drivewise & Arity — Spinning Off the Data Subsidiary

Allstate Drivewise launched in 2010. In 2016, Allstate carved out its telematics data assets into a separate subsidiary, Arity. By 2026 Arity has grown into a B2B vendor that supplies data and scoring models not only to Allstate but also to GEICO, Nationwide, and others.

Drivewise's differentiator is immediate post-trip feedback. As soon as a trip ends, the app shows that trip's score and risk events. Accumulating high-score trips yields up to 40% off at renewal.

Arity collects data from auto OEMs, navigation apps, and ride-share companies, building a driver behavior dataset that covers roughly 75% of US drivers. As of 2026 the dataset's monetized value is comparable to Allstate's own insurance revenue.

// Arity-style end-of-trip score — simplified weighted-sum model
type Trip = {
  miles: number
  durationMin: number
  harshBrakes: number
  harshAccels: number
  phoneSeconds: number  // screen-on time while moving
  speedingMiles: number // miles spent > posted limit + 10mph
  nightMinutes: number  // minutes between 00:00 and 04:00
}

function tripScore(t: Trip): number {
  // base 100, subtract penalty per event normalized by mileage
  const m = Math.max(t.miles, 0.1)
  const brakeP = (t.harshBrakes / m) * 12
  const accelP = (t.harshAccels / m) * 8
  const phoneP = (t.phoneSeconds / Math.max(t.durationMin * 60, 1)) * 40
  const speedP = (t.speedingMiles / m) * 25
  const nightP = (t.nightMinutes / Math.max(t.durationMin, 1)) * 10
  const raw = 100 - (brakeP + accelP + phoneP + speedP + nightP)
  return Math.max(0, Math.min(100, raw))
}

Chapter 5 · State Farm Drive Safe & Save — The Most Conservative PHYD

State Farm has the largest US auto insurance share (about 17%) but has been conservative on UBI. Drive Safe & Save (DSS) launched in 2011 but initially focused on Ford Sync and OnStar integrations. In 2018 they shipped a smartphone app and shifted to a true PHYD model.

Two DSS policies stand out. First, no rate increases. A low score will not push you above the base rate. Second, the average discount is around 8% with a 30% cap, more conservative than Progressive. The trade-off is a low-friction onboarding, and more than 70% of new auto policies opt into DSS.

State Farm is also conservative in the scoring algorithm itself. Mileage carries the largest weight, and harsh braking is secondary. The model favors low-mileage drivers.

Chapter 6 · Liberty Mutual RightTrack — A 90-Day Trial

Liberty Mutual RightTrack uses a 90-day trial model. You get a 5% discount immediately at enrollment, and up to 30% more after 90 days of monitoring. The differentiator is that data collection stops after the trial. "Be evaluated once and then be free" became a marketing point that resonated.

Technically RightTrack licenses CMT's SDK and tunes the scoring model in-house. The data retention period is just 90 days — making this one of the most aggressive applications of GDPR-style data minimization in the US market.

Chapter 7 · Root Insurance — Telematics-First, Chasing Profitability Post-IPO

Root Insurance, founded in 2015 in Ohio, is a telematics-first carrier. "We only accept people who drive well" is the slogan — enrollment itself requires passing a telematics evaluation.

The onboarding flow is distinctive. After installing the app you go through a 2–4 week Test Drive period. If your driving score is low during that window, the application is rejected. If you pass, the rate is set by score band.

Root was unprofitable for years after its 2020 IPO (a -$298M result in 2022). Loss ratios improved through 2024, the company turned profitable in 2025, and in 2026 it operates in 35 states including California and Texas. The hypothesis that loss ratios drop as telematics data accumulates has been partially validated.

# Root-style enrollment evaluation — accept/reject + rate band from Test Drive period
def evaluate_test_drive(trips: list[dict]) -> dict:
    """trips: list of dicts with miles, harsh_brake_count, phone_usage_sec, etc.
    Returns decision: 'accept' / 'reject' and rate band 1-10."""
    total_miles = sum(t['miles'] for t in trips)
    if total_miles < 300:
        return {'decision': 'pending', 'reason': 'insufficient_mileage'}

    # weighted composite score
    brake_per_100mi = sum(t['harsh_brake_count'] for t in trips) / total_miles * 100
    phone_pct = sum(t['phone_usage_sec'] for t in trips) / sum(t['duration_sec'] for t in trips)
    speed_pct = sum(t['speeding_miles'] for t in trips) / total_miles

    score = 100 - (brake_per_100mi * 2.5 + phone_pct * 150 + speed_pct * 100)
    score = max(0, min(100, score))

    if score < 35:
        return {'decision': 'reject', 'score': score, 'reason': 'high_risk'}

    # rate band: 1 (cheapest) to 10 (most expensive accepted)
    band = max(1, min(10, int(11 - score / 10)))
    return {'decision': 'accept', 'score': score, 'rate_band': band}

Chapter 8 · Metromile — Life After the Lemonade Acquisition

Metromile, launched in 2011, was a PAYD specialist. The model: a small monthly base (around 29 USD) plus 6–12 cents per mile. Attractive to low-mileage urban drivers.

In 2022 Lemonade acquired Metromile, and as of 2026 Metromile operates as the PAYD line of Lemonade Auto. Its standalone brand presence weakened, but the per-mile pricing model has resurged as a beneficiary of permanent remote work — drivers under 200 miles per week find traditional insurance overpriced.

Technically Metromile used an OBD-II dongle (Metromile Pulse). Post-Lemonade integration a smartphone mode was added, but the dongle channel is still operated.

Chapter 9 · Cambridge Mobile Telematics (CMT) — The B2B Power Behind the Carriers

Cambridge Mobile Telematics, founded in 2010 by MIT alumni, is a B2B telematics company. It does not sell its own insurance — it supplies the DriveWell platform to insurers and auto OEMs.

As of 2026, CMT counts more than 100 insurers and fleet operators in 25+ countries as customers. The back end of Liberty Mutual RightTrack, GEICO DriveEasy, Travelers IntelliDrive, parts of Carrot Permile, and Japan's SOMPO Drive Save all runs on CMT.

CMT's distinguishing feature is crash detection and post-incident handling. When patterns suggestive of a crash (multi-axis acceleration above 0.6g) are detected, the system automatically triggers a 911 call or a carrier-agent call. That single capability moved CMT from a scoring vendor into a driver-safety platform.

Chapter 10 · Octo Telematics — The Other Giant, Built in Europe

Octo Telematics, founded in Italy in 2002, is the European telematics heavyweight. By 2026 Octo holds 6 million cumulative policies and partnerships with 50 carriers. In Europe Octo has a higher share than CMT.

Octo's strength is its OBD/embedded device + smartphone hybrid. In Italy and the UK, certain insurance lines are mandated by anti-theft laws to use OBD-style telematics, and Octo dominates that market. Acquired in 2024 by Modus, Octo is now accelerating global expansion.

Within the EU's GDPR regime, Octo became a reference for separation of personally identifiable information (PII). Location data and score data are isolated into distinct stores; carriers receive only scores and never the raw location.

Chapter 11 · Tesla Insurance — The Vehicle Itself Is the Telematics

Tesla Insurance launched in California in 2019. As of 2026 it operates in 12 states including California, Texas, Illinois, Colorado, Arizona, Virginia, and Maryland (four new states added during 2024–2025).

Tesla Insurance's real edge is that the car itself is the telematics device. Autopilot data, FSD (Full Self-Driving) data, and onboard camera and radar data all feed into the insurance system. The premium is recalculated monthly via a 0–100 Real-Time Safety Score.

The five core inputs of Safety Score:

Forward Collision Warnings per 1,000 mi — Rate of forward-collision warnings.
Hard Braking — Deceleration above 0.3g (same threshold as Progressive).
Aggressive Turning — Lateral acceleration above 0.4g.
Unsafe Following — Headway under one second.
Forced Autopilot Disengagement — Situations where Autopilot was forced to disengage (e.g., inattention).

Tesla claims that drivers scoring above 95 have roughly one-third the accident rate of the average. If that holds, Tesla Insurance's loss ratio should be far below traditional carriers. That said, 2024–2025 reporting showed loss ratios above 200% in several states, and criticism of the pricing model's limits has grown in step.

Chapter 12 · Korea — The Market Carrot Permile Built

Korea's UBI market was effectively created by Carrot General Insurance's Permile. Launched in 2020, Carrot Permile is a PAYD model with a per-km variable rate of about KRW 32 plus a small monthly base. It started on an OBD-II dongle and added smartphone mode in 2023.

In 2025 Carrot crossed 1 million cumulative subscribers, and in 2026 the IPO process is now well underway. With roughly 12% share of Korea's direct auto insurance market, Carrot earned a major position in just six years.

The incumbents followed. KB Direct Drive (KB Insurance) is a PHYD line offering up to 15% off based on a safe-driving score. Samsung Fire Direct MyCar combines mileage with driving habits. Hyundai Marine HiCar also runs telematics discounts on some lines.

A specific feature of Korea is the stringency of consent under PIPA (the Personal Information Protection Act). Separate consent for location data is mandatory and enrollment screens are long. Carrot addresses this by promising to store location data only as vehicle-level aggregates, without mapping it to drivers — a promise that became central to its marketing.

# Carrot-style per-km premium calculation — base + variable rate
def monthly_premium_carrot(km_driven: float,
                            base_monthly_krw: float = 12000,
                            per_km_krw: float = 32.0,
                            night_km: float = 0.0,
                            night_surcharge_pct: float = 0.10) -> float:
    """Simplified Carrot Permile — monthly premium (KRW) with a 10% nighttime surcharge."""
    base = base_monthly_krw
    variable = km_driven * per_km_krw
    night_extra = night_km * per_km_krw * night_surcharge_pct
    total = base + variable + night_extra
    return round(total)

Chapter 13 · Japan — The SOMPO, Tokio Marine, and Sompo Japan Trio

Japan's UBI market started later than the US or Korea but has grown rapidly since 2021.

SOMPO Drive Save, launched in 2021, hit its five-year mark in 2026. It runs on the CMT back end and offers up to 20% off for safe-driving scores at or above 80. It has crossed 500,000 cumulative subscribers.

Tokio Marine & Nichido Drive Agent Personal combines a drive recorder with telematics. After an accident, the drive-recorder video is automatically uploaded to the carrier. The attached-device model is more accurate but more expensive.

Sompo Japan Smiling Road specializes in driver coaching. When the score is low, the app gives concrete improvement guidance like "brake a bit earlier next time." It positions telematics as a driver-coaching tool rather than a pure discount mechanism.

A particular feature of Japan is the existence of a separate line for monitoring of elderly drivers. Options that let family members receive telematics alerts for drivers over 75 are popular.

Chapter 14 · The Scoring Algorithm in Practice — Harsh Braking, Cornering, Phone Use

Let's look at the four key events more carefully.

Hard Braking — A 0.3g longitudinal deceleration is a common threshold. Decelerating gradually beats slamming the pedal.
Hard Acceleration — Longitudinal acceleration above 0.35g. Catches the floor-it-on-green pattern.
Hard Cornering — Lateral acceleration above 0.4g. Catches sharp turns and lane changes.
Phone Usage — Time with the phone screen on while the vehicle is moving (above 8 mph). Measured via iOS ScreenTime API and Android UsageStats API.

The core problem in all of this is GPS and accelerometer noise. In cities, GPS reflects off buildings. The accelerometer axes depend on the phone's pose. The standard pipeline is to Kalman-filter the position to smooth it, and to rotate axes via PCA to convert phone-frame acceleration into vehicle-frame acceleration.

# Simplified 1D Kalman filter — smoothing GPS speed estimates
class KalmanSpeed1D:
    def __init__(self, process_var: float = 0.5, measure_var: float = 4.0):
        self.x = 0.0          # estimated speed (m/s)
        self.P = 1.0          # estimate covariance
        self.Q = process_var  # process noise
        self.R = measure_var  # measurement noise

    def step(self, gps_speed: float, dt: float) -> float:
        # predict: constant-velocity model, no control input
        self.P = self.P + self.Q * dt
        # update with GPS observation
        K = self.P / (self.P + self.R)
        self.x = self.x + K * (gps_speed - self.x)
        self.P = (1 - K) * self.P
        return self.x

# usage
kf = KalmanSpeed1D()
for sample in gps_stream:
    smooth = kf.step(sample.speed_mps, sample.dt)
    if smooth > 30.0:  # > 30 m/s = > 67 mph
        flag_speeding(sample.t)

Chapter 15 · The Diagnostic Standard — OBD-II and OBD-III(?)

OBD-II is the vehicle diagnostic standard mandated in the US since 1996 (gasoline in 2001 and diesel in 2004 in the EU). A 16-pin diagnostic port exposes engine, transmission, and emissions information.

UBI devices read the following from OBD-II:

PID 0x0D — Vehicle Speed: the speed reported by the vehicle itself (more accurate than GPS).
PID 0x0C — Engine RPM: acceleration and gear-change patterns.
PID 0x11 — Throttle Position: accelerator pedal position.
PID 0x42 — Battery Voltage: confirms the device is powered by the vehicle.

OBD-II is a 30-year-old standard, and discussions around OBD-III (a cellular-direct emissions reporting protocol) are ongoing in 2026, though stalled by privacy concerns.

EVs are an awkward fit for OBD-II. Tesla uses its own diagnostic port, and OBD-II-compatible devices don't connect. OBD-II-based telematics is gradually obsolescing in the EV era as a result.

Chapter 16 · The Accuracy Limits of Smartphone Telematics

Smartphone telematics accuracy is roughly 80–90% of OBD-II. The main error sources:

Driver Attribution — You don't know who was driving. If a parent's car is driven aggressively by their teen, the parent's score suffers. CMT's BLE beacon is an attempt to solve this.
Phone Pose — Whether the phone sits in a cup holder, a pocket, or a dash mount changes the accelerometer axes.
GPS Drift — Cities, tunnels, and under-bridge stretches make GPS jump.
Battery Saver / Background Restriction — Android's Doze mode and iOS Background App Refresh restrictions cause data drops.

The standard mitigations are: estimate vehicle axes from accelerometer PCA, fuse Doppler GPS speed with acceleration, and use the OS ActivityRecognition API ("In-Vehicle" classification) to start and end driving sessions. CMT, Arity, and Sentiance (an Allianz subsidiary) lead the technology in this space.

UBI handles some of the most sensitive personal data — location, driving time, acceleration patterns, phone-use duration. All of it is behaviorally identifying.

Under GDPR (EU), UBI data must satisfy:

Explicit consent (Article 7) — Telematics processing requires separate consent from the insurance contract itself.
Purpose limitation (Article 5(1)(b)) — Cannot be used outside premium calculation (e.g., for ads).
Minimization (Article 5(1)(c)) — Only data necessary for the score.
Storage limitation (Article 5(1)(e)) — Deleted after the insurance period plus the mandatory retention window.
Right against automated decision-making (Article 22) — Right to demand human review.

During 2024–2025, GDPR fines against telematics carriers in Italy, Germany, and France accumulated to more than EUR 5M. The dominant violation: "location retention without consent."

CCPA (California) is weaker than GDPR but provides a "Do Not Sell My Personal Information" right. Tesla Insurance gets explicit consent for driving data at enrollment, but withdrawing that consent may cause renewal to be denied.

Korea's PIPA mandates separate consent for location data. Carrot and KB Direct both have dedicated consent boxes at enrollment with stated retention periods.

Chapter 18 · Discrimination Concerns — Who Loses Under UBI

UBI advocates argue, "We look at behavior, not at identity." Critics point to disparate impact:

Night-driving penalty — Penalizes healthcare and service workers on night shifts.
Long-commute penalty — Penalizes outer-suburb and lower-income residents.
City-driving penalty — Dense urban traffic and frequent signals naturally cause hard-braking events.
Data scarcity — Drivers who don't drive much during a trial cannot earn a score and cannot earn a discount.

Several US states (California, New Jersey, Massachusetts) have tightened insurance-department review of UBI models. In 2024 the California Department of Insurance issued guidance requiring annual reporting on whether scoring models exhibit a "de facto discriminatory effect" on race or income.

Chapter 19 · Crash Detection and Emergency Calls — Telematics' Real Value

Telematics premium discounts are a marketing tool, but the real utility may be crash detection and emergency calls.

CMT reports roughly 95% crash-detection accuracy. When a crash is detected, the system pings the user within 30 seconds and, with no response, automatically calls 911 or pre-registered emergency contacts. Since 2022, CMT-based systems have triggered more than 50,000 emergency calls in the US.

Apple built its own crash detection into iOS from the iPhone 14 (2022); Google Pixel has had it since the Pixel 4. This reduces carrier dependency but carriers leverage the crash data to automate claims — a flow where an insurance quote arrives within an hour of an accident became routine in 2026.

Chapter 20 · Fleet Telematics — The Other Market, Driven by Geotab and Samsara

A separate, very large market for fleet telematics runs alongside consumer UBI. It is the tooling shipping companies, rental fleets, and ride-share operators use to manage their own vehicles.

Geotab, headquartered in Canada, manages more than 4 million vehicles. OBD-II + cellular devices track location, fuel, and driving behavior.

Samsara, founded in 2015 and IPO'd in 2021, is the US heavyweight. Devices, the cloud platform, and AI cameras are integrated. ARR is around USD 1.4B as of 2026 — the leader in fleet SaaS.

Verizon Connect (formerly Fleetmatics), Lytx, and Motive (formerly KeepTruckin) follow. In Korea, SK Telecom T-Map Biz and KT GiGA Fleet; in Japan, NTT Docomo Biz lead.

Fleet telematics is separate from insurance, but some fleet policies accept Samsara/Geotab data as a basis for premium discounts. For trucking companies it means "DOT compliance and insurance discount" can be solved at once.

Chapter 21 · UBI Pricing in Practice — Discount Ranges and Loss Ratios

Average UBI discounts vary by carrier, but in the US:

Progressive Snapshot — 17% average, 30% max (about 20% see rate increases).
Allstate Drivewise — Around 15% average, 40% max.
State Farm DSS — Around 8% average, 30% max, no rate increases.
Liberty Mutual RightTrack — 5% at enrollment + up to 30% after 90 days.
GEICO DriveEasy — Around 13% average, 25% max.
Tesla Insurance — Safety scores 80+ can yield 25% or more in savings.

On the loss ratio side, UBI subscribers' accident rates are reportedly 30–40% lower than the general population. That said, this includes self-selection bias — safer drivers opt into UBI voluntarily. Carriers like Root that gate enrollment on telematics report lower loss ratios still, but they trade away addressable market at the door.

Chapter 22 · UBI and EVs — A New Scoring Model

EVs have brought two shifts to UBI.

First, regenerative braking has become standard, meaning brake-pedal-use patterns differ from ICE vehicles. Tesla's "Hold" mode stops the car when you lift off the pedal. As a result, brake-event detection must be accelerometer-based, not pedal-based.

Second, charging patterns can become UBI inputs. Tesla Insurance says it does not use charge frequency, charger type, or DC fast-charging ratio as score inputs, but some OEM-partner insurance lines treat charging patterns as a secondary signal for location data.

Third, the new reality: EVs are more expensive to insure than ICE cars. Repair costs (especially batteries) are higher, and even with LDW/AEB the loss amount per claim is larger. Telematics discounts therefore work at a larger absolute scale for EV drivers — 15% off a high premium is a lot in dollars.

Chapter 23 · Autonomy and UBI — Who Was Driving?

As Level 2 and 3 autonomy spread, UBI must answer a new question: when an accident happens in autonomous mode, who gets the score?

Tesla logs Autopilot-engaged time as a separate category. Hard braking that happens during Autopilot is excluded from the driver's score and reclassified as validation data for the OEM's autonomy stack. The policy favors drivers and simultaneously incentivizes Tesla to gather richer autonomy data.

Fully autonomous ride-share operators like Waymo and Cruise carry commercial autonomous-driving liability insurance instead of regular auto policies. Munich Re and Swiss Re are the major reinsurers here. In that model UBI as a concept loses meaning — there is no driver.

Chapter 24 · The 2026 Trend Lines — Where We're Going

To close, five flow lines for 2026.

Embedded telematics goes mainstream — Tesla, BMW, Hyundai, and Toyota expand OEM insurance lines. Smartphone telematics gradually moves to a secondary channel.
Data marketplaces — Following the Arity model, more carriers sell telematics data B2B.
AI post-incident analysis — Tools that use AI to analyze crash video and automatically assign fault percentages settle inside carriers.
Tighter privacy regulation — The EU AI Act's implementation makes human review of automated scoring more mandatory.
EV and autonomy insurance branch off — A dedicated line splits from general UBI for EVs and autonomous vehicles.

The next ten years of UBI head not into "the era of driver data" but into "the era of vehicle data." Drivers will matter less; the vehicle itself, and the OEM, will be the subject of the score.

Epilogue — "You Drove Well"

Back to Chicago. Six months later that driver gets a Progressive renewal notice: "Your Snapshot score is 87. Your renewal will apply a 22% discount."

22% trims about 380 dollars off their annual premium. They are satisfied. Progressive is satisfied too — the data show their accident rate is about 60% of the average policyholder. The seat where those two satisfactions meet is the seat that telematics + UBI occupies in 2026.

But this seat has a shadow. Night-shift workers. City dwellers. People who refused data consent. Parents and teens whose scores are not separated when a phone moves between drivers. UBI's efficiency and its discrimination are two sides of the same coin.

2026 auto insurance is growing both the light and the shadow together. And the next decade is moving toward an era in which the vehicle itself, not the person, is the subject of the score.