BRENNEN

Industries / Finance and business services / Insurance / InsurTech

Cover priced by use

Insurance priced by how much or how well a customer drives, or how they live, as measured by a device in the car, a phone app or a health program.

7

Milestones achieved, from 2006 to 2026.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 2006

    Pay as you drive, paused

    Norwich Union, now Aviva, launched a "pay as you drive" car policy in the UK in October 2006, priced on data from a device in the car. In June 2008 it said it would "pause" the scheme. It had aimed for 100,000 customers, and reports suggested about 10,000 had signed up, Insurance Times wrote. A spokesman said the technology "has not evolved as much as we thought it would". Graeme Trudgill of the British Insurance Brokers' Association said: "The customers don't like the whole Big Brother attitude."

    LessonDrivers had to agree to be watched, and few did.

    transportxtra.cominsurancetimes.co.uk

    Credit

    Norwich Union

  2. 2011

    Rewards for safe driving

    Discovery Insure launched in South Africa in 2011 and calls itself "the first telematics-based vehicle insurer". Telematics means sending data about how a vehicle is used from a device or a phone. Its Vitality Drive program scores habits including "avoiding cellphone use" and pays back up to 50% of a driver's fuel spending each month. Discovery reports a 24% reduction in accident frequency among its clients, with no outside check.

    discovery.co.za

    Credit

    Discovery, by its own account

  3. 2018

    Life cover tied to habits

    John Hancock's Vitality program lets customers lower their premiums by earning points "for health-related activities such as exercising, getting a flu shot and logging their annual visit to the doctor". In September 2018 California's Department of Insurance announced that John Hancock "will stop selling traditional life insurance policies in California". Commissioner Dave Jones said the department had worked "to expedite approval" of the new product.

    insurance.ca.gov

    Credit

    John Hancock; California Department of Insurance

  4. 2022

    Per-mile insurer sold cheaply

    Metromile, founded in 2011, charged drivers by the mile. It listed in February 2021 by merging with a special purpose acquisition company (SPAC), a listed shell company, at an equity value of $1.3 billion. In November 2021 Lemonade agreed to buy it at about $500 million. When the deal closed in July 2022, Lemonade said it paid "under $145 million in stock" for a company holding "over $155m in cash" and data from 500 million car trips.

    LessonCharging by the mile did not by itself build a business worth more than its cash; the buyer named the cash and the trip data as what it gained.

    techcrunch.comsec.govreinsurancene.wslemonade.com

    Credit

    Metromile; Lemonade

  5. 2025

    Prices can rise too

    The Maryland Insurance Administration surveyed car insurers on their telematics programs. It found 13.23% of policies enrolled. Of enrolled policies, 23.6%, or 62,233, had their price raised at renewal, and 31.16% had it lowered.

    LessonA program sold as a way to save can also be used to charge more.

    insurance.maryland.gov

    Credit

    Maryland Insurance Administration

  6. 2026

    Driving data sales banned

    According to the US Federal Trade Commission (FTC), General Motors collected drivers' locations and driving behavior through OnStar, "sometimes as often as every three seconds". GM shared the data with Verisk from 2016 and LexisNexis from 2019, and both used it to compile consumer reports. GM ended the deals in March 2024, and Verisk has since stopped selling such reports. The FTC proposed an order in January 2025 and finalized it in January 2026, barring GM for five years from sharing such data with consumer reporting agencies. In January 2025 Texas sued Allstate and its data company Arity, alleging they collected location data "from over 45 million consumers nationwide".

    LessonDriving data shared for insurance pricing without clear consent led to a five-year ban.

    ftc.govftc.govftc.govtherecord.mediainsurancenewsnet.com

    Credit

    US Federal Trade Commission; Texas Attorney General

  7. 2026

    Italy's black boxes slip

    IVASS, the Italian insurance regulator, reported that 16% of car insurance contracts included a "scatola nera", a black box that records driving, in the first quarter of 2026, against 17% a year earlier. The share was 54.6% in the province of Caserta and 44% in Naples.

    ivass.it

    Credit

    IVASS

Watch for

  1. Cheaper miles with self-driving

    Lemonade says its car insurance cuts per-mile rates by roughly 50% for miles driven with Tesla's Full Self-Driving system engaged, from 26 January 2026 in Arizona and a month later in Oregon. The lower risk comes from Lemonade's own models.

    Achieved whenloss figures or an outside study show fewer or smaller claims on those miles.

    claimsjournal.com

    Claimed by

    Lemonade, January 2026

  2. Your own driving report

    Arity said it would make it "easier for individuals to check whether a driving report exists for them, and view it directly".

    Achieved whenthe tool is live and an outside party, such as a consumer group, confirms that drivers can get their reports through it.

    repairerdrivennews.com

    Claimed by

    Arity, June 2025

Open

  1. 01

    Telematics terms in public

    A free table of the usage-based programs of the ten largest US car insurers in ten states. For each it gives the largest discount, the largest surcharge, whether the price can rise at renewal, and what is measured (miles, braking, speed, time of day, phone use). Every cell links to the page of the state rate filing it came from.

    The brief
    Credit · unclaimed
  2. 02

    Phone sensors against a reference

    An open dataset of at least 200 drives in which a phone's motion sensors are recorded alongside a reference unit fixed to the car and a dashcam. Hard braking and phone handling are labelled from the video, and the dataset shows how well a simple detector on the phone data matches the reference. Planned hard stops are made only on a closed private lot. Phone handling is done only by a passenger, never by the driver.

    The brief
    Credit · unclaimed

Open 01 · the brief

Telematics terms in public

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A free table of the usage-based programs of the ten largest US car insurers in ten states. For each it gives the largest discount, the largest surcharge, whether the price can rise at renewal, and what is measured (miles, braking, speed, time of day, phone use). Every cell links to the page of the state rate filing it came from.

Why it is still open

In January 2025 the Consumer Federation of America told the NAIC there is "a severe lack of reliable information about telematics" and asked for data "broken out by company, for each of the nation's ten largest insurers". Maryland's 2025 survey combines its insurers and does not name them.

consumerfed.orginsurance.maryland.gov

Built with what exists

SERFF Filing Access and state insurance department portals, used within each site's terms; pdftotext; a language model such as Claude to read the filings and cite each page (see the Note "Seven more flagged"); Python and pandas.

  1. First step

    Find one large insurer's telematics filing for Maryland and record the five fields by hand.

  2. Done looks like

    100 insurer-and-state pairs filled in, every value traced to a filing page. A second person checks 50 values picked at random and agrees with at least 48. The method is published beside the table.

Open 02 · the brief

Phone sensors against a reference

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open dataset of at least 200 drives in which a phone's motion sensors are recorded alongside a reference unit fixed to the car and a dashcam. Hard braking and phone handling are labelled from the video, and the dataset shows how well a simple detector on the phone data matches the reference. Planned hard stops are made only on a closed private lot. Phone handling is done only by a passenger, never by the driver.

Why it is still open

The one comparison found is a 2012 study of 78 drivers. It set an iPhone 4 against a motion unit plugged into the car, found that the phone measurements "tend to overestimate critical driving events", and did not test phone use. Cambridge Mobile Telematics, a vendor, reported to the NAIC that the accuracy of its phone-distraction detection rests on user feedback: "only 2.2% of events users provided feedback about were found to be non-distraction events".

alexandria.unisg.chcontent.naic.org

Built with what exists

a phone sensor logging app such as phyphox; an OBD-II data logger and a GNSS and motion unit fixed to the car; a dashcam; Label Studio for video labels; Python.

  1. First step

    Record five drives with a phone and an OBD-II logger at the same time and line up their clocks.

  2. Done looks like

    200 drives by at least 10 drivers on at least 3 phone models, every labelled event time-stamped. The braking measured by the phone is compared with the reference, and the share of passenger phone-handling events caught is reported. Everything is published under an open license.