Couple of posts from the Smart 421 people on usage-based car insurance.
Their conclusion: it isn't going to save anyone enough money to justify the cost and inconvenience and to overcome the concern for privacy. Adoption only becomes viable when the devices become a commodity (multiple providers, standard interface, routinely fitted to new cars), with a strong carrot/stick from the government (e..g legislation or road charging).
This is pretty close to what I said when the Norwich Union experiment was abandoned last year (Pay as you drive 2).
The Smart 421 bloggers also have some comparisons with Pay-As-You-Go models for other utilities, including domestic energy and water, and telecoms. Telecoms is clearly more complex than energy and water consumption, and has much more complicated pricing schemes. What are the lessons for software pricing - e.g. Anything-as-a-Service?