Usage-Based Insurance Statistics 2026

A driver enrols in a telematics programme in March, drives carefully all summer, and opens a renewal in September showing a higher premium than last year. Nothing has gone wrong. The rate moved for reasons that have nothing to do with the score, and no one told them that could happen.

That gap between what the customer expects and what the product does is where most usage-based programmes lose people.

The measurement side works. Mileage, time of day and driving behaviour give a motor insurer a far more direct view of exposure than a postcode and a birth date. The commercial question is whether the proposition survives the rest of the policy lifecycle.

These usage-based insurance statistics cover adoption, pricing, customer experience and the openings in specialty motor. Historical estimates, forecasts and current product terms are labelled separately, because the three get mixed together constantly.

Research updated 9 September 2026. Reporting periods are stated alongside the figures.

Usage-based insurance statistics at a glance

Policy counts measure scale, satisfaction scores measure experience, and a discount reported by one insurer describes one customer group. Each answers a different question, and none of them is an industry average.

What is usage-based insurance?

Usage-based insurance prices motor risk from information about how a vehicle is actually used. Telematics collects it through a smartphone, a device fitted to the car, or the vehicle’s own connected system, and the NAIC identifies mileage, driving behaviour and the timing of journeys among the relevant inputs.

There are three broad ways to turn that into a product.

They overlap in practice. A mileage product can also reward behaviour, and an insurer can use a short observation window to set a renewal adjustment rather than repricing continuously.

Where UBI fits in specialty motor

Using telematics does not by itself make a policy a specialty or surplus lines product. The specialty angle appears when the underlying exposure is unusual or moves around: a courier who only works weekends, a fleet with a distinctive operating pattern, a low-mileage vehicle used intensively for three weeks a year.

Zego’s UK pay-as-you-go delivery cover shows how precise that gets. Its published charging model applies a one-hour minimum per session and then charges by the minute, so several short sessions can bill more hours than the driver actually worked.

The insurer discloses that openly, which is the right call, but it is also the design problem in miniature. The unit of exposure might be a mile, a shift or a score. The policy and the invoice still have to explain what the customer bought.

Telematics adoption: 34 million policies, very unevenly spread

Berg Insight’s ninth-edition research counts about 34 million insurance telematics policies across Europe and North America at the end of 2024. Italy accounts for roughly 10 million of them, Germany 1.5 million and the UK 1.3 million.

Italy is not a rounding difference. It is most of a continent.

The same research forecasts 20.1 million European and 31.8 million North American policies by 2029, growing at 7.8% and 9.5% a year respectively. Those are forecasts, not observed 2026 counts.

Concentration like that should temper any transplant. A proposition refined in a market where telematics is decades old carries assumptions about distribution, onboarding and customer familiarity that do not travel.

These figures also count policies, not app downloads, connected vehicles or unique drivers. A connected car is not enrolled in anything. An installed app may never capture enough usable trips to price from.

Measuring your own adoption works better as a sequence: eligible customers, offers made, enrolments, successful activation, sufficient driving data, and still participating at renewal. Stopping at the sign-up hides everything that goes wrong afterwards.

How much can drivers save with usage-based insurance?

Progressive says Snapshot drivers who save do so by an average of $328 a year. The same page says about two in ten participants get a rate increase, and that the earned discount normally arrives at renewal, once the programme has collected enough driving data.

The qualifier carries the whole sentence. That is the average among drivers who save, not the average across everyone who enrolled.

For a product team the distribution matters more than the headline. How many customers receive a discount, how many an increase, how large each adjustment is, and how many leave before there is enough data to calculate one at all.

A pricing model can be technically sound and still disappoint, if the opening offer set an expectation the renewal cannot meet.

Dynamic pricing: separate the rate from the bill

Two different changes get bundled together under dynamic pricing in insurance: a change in how much insured activity there was, and a change in the price attached to that activity.

Nationwide’s SmartMiles is the first kind. Its monthly cost combines a base rate with a variable mileage component, so a busy month costs more even when the per-mile rate has not moved at all.

A behaviour-based programme is the second kind. An observation period produces a score, the score feeds a later premium adjustment, and the customer sees feedback in between.

Three things should be explicit in the design: which miles, activities or signals count; how they affect the premium; and when the revised amount shows up on the bill. Pricing, service and billing teams then work from one explanation, which is what makes a dispute answerable. Was the trip attributed wrongly, was the rate applied wrongly, or did the customer expect the adjustment sooner?

Road-trip allowances need their own measure

Low-mileage drivers still take holidays. SmartMiles handles it with a road-trip exception under which only the first 250 miles in a single day count toward mileage charges, subject to the product’s terms and availability.

That is a billing feature, and it is worth keeping separate from claims under roadside assistance or rental reimbursement. A daily mileage cap says nothing about how often assistance gets used or what a breakdown costs the insurer.

Useful measures here would be the share of customers hitting the cap, its premium effect, and retention among occasional long-distance drivers. Those are internal metrics; the reviewed sources establish no industry benchmark for them.

Customer experience decides whether telematics keeps its appeal

Collecting driving data is the easy half. Customers also have to trust the score that comes back.

JD Power’s January 2026 outlook, drawing on its 2025 studies, puts the insurance app 75 points below the vehicle’s own system. The comparison shows an association rather than a cause: customer mix, insurer service and programme design all differ between those groups too.

In its separate 2026 US Auto Insurance Study, fielded to 52,216 customers between April 2025 and April 2026, JD Power ranked Nationwide highest for UBI for the third consecutive year.

“The market has clearly shifted from a pricing crisis to an experience challenge”
Stephen Crewdson, managing director of insurance business intelligence, JD Power

Crewdson was describing the wider auto market, and it lands hardest here. A customer querying a passenger trip logged as their own driving should get an answer without explaining it three times to three channels.

Driving data can support prevention as well as pricing

Telematics is not only a way of finding the lower-risk customer at inception. The same data flags behaviour worth feeding back.

A September 2025 report from the Governors Highway Safety Association and Cambridge Mobile Telematics links high rates of hard braking with 103% higher expected losses, and excessive speeding with 71% higher predicted losses. Those are risk-model findings, not measured savings from enrolling anyone in a programme.

The cost side is moving in the same direction. LexisNexis Risk Solutions reported that bodily injury accounted for more than 26% of total auto claims dollars in 2025, against under 20% in 2022 in its analysed data.

Which makes the chain worth testing end to end. Does feedback change behaviour, does the change persist, and does it reach claims experience once exposure and selection are allowed for? A safer enrolled portfolio partly reflects who chose to join, and separating that from genuine improvement is what makes the business case honest.

Consent and data quality are part of the product

In January 2026 the FTC finalised an order settling allegations that GM and OnStar collected and sold geolocation and driving data without proper consent. It carries a five-year restriction on sharing covered driver data with consumer reporting agencies, alongside consent and consumer-control requirements.

The order names two companies. It is not a ruling on usage-based insurance.

What it does establish is the standard a programme will be judged against: clear enrolment choices, documented permissions and a route to correct the record. A driving score is only useful when its source, its permitted uses and the identity of the driver behind it are all reliable.

What insurers should measure next

The useful measures pair adoption with commercial and customer outcomes:

• Activation. How many enrolled customers produce enough valid data to participate at all.
• Pricing outcomes. The share receiving a discount, no change or an increase, with the size of each.
• Retention by score, by pricing outcome and by collection method.
• Claims performance by exposure, against a comparison group that did not enrol.
• Service friction. Disputed trips, missing data, score queries and how long each takes to resolve.
• Programme economics. Data, servicing and acquisition costs set against the underwriting result.

Together they answer the only question that matters at renewal: expand this, redesign it for a narrower niche, or simplify an experience customers are struggling with.

None of them sit naturally in a system built around an annual premium and a fixed rating table. Mileage, scores, observation windows and mid-term adjustments all have to reach the billing system in a form the customer can read on an invoice.

Openkoda gives insurers and MGAs configurable products and pricing over policy, claims, billing and reporting, which is the part of a UBI programme that is easy to underestimate and expensive to retrofit. The same pattern shows up in embedded car insurance, where the exposure arrives from someone else’s platform.