A Texas auto insurance pool repriced three ride-share drivers' telematics scores mid-term, triggering premium jumps of roughly 40%. The increases, totaling an estimated $2,700 over six months across the three policies, traced back to a single decision by their insurer: a mid-term repricing of telematics scores within a state-regulated auto insurance pool. The pool, designed to spread risk among high-risk drivers, had never before allowed an insurer to re-score a policy mid-term based on a proprietary algorithm. The drivers—let's call them Carlos M., Elena R., and David K.—used their personal vehicles for ride-share work and saw their telematics scores drop from the mid-80s to the low 60s in two months. The insurer claimed the algorithm reflected higher risk from ride-share trips, but the pool's actuaries disagreed on the correlation. A Texas Department of Insurance inquiry closed without finding, in part because the regulator lacked access to the algorithm's inner workings. The case illustrates a growing tension in usage-based insurance: when a telematics score is not just a measure of driving behavior but a hybrid of personal and commercial trip data, who decides what the score means, and who gets to see how it's calculated?

The Single Pool That Priced Three Ride-Share Telematics Scores

The Texas auto insurance pool in question operates as a shared market for drivers who cannot easily obtain coverage in the standard market. Insurers that participate in the pool agree to accept a share of the pool's policies and to use the pool's rating structure. But in this case, one insurer introduced a telematics program within the pool that allowed it to re-score drivers mid-term based on data from the insurer's own mobile app. The three drivers, all of whom had signed up for the telematics discount, received notices in February 2024 that their scores had been updated. The update reflected a blend of personal trips and ride-share trips logged through a platform-integrated app. The insurer did not notify the drivers of the score change until the renewal statement arrived.

The pool's rate filing with the Texas Department of Insurance had not disclosed that telematics scores would be weighted by trip type. Pool actuaries, when later briefed, argued that the rating structure assumed telematics scores reflected only driver behavior—hard braking, rapid acceleration, mileage—not the inherent risk profile of ride-share versus personal driving. The insurer's proprietary algorithm, however, assigned a higher risk factor to ride-share trips based on time-of-day patterns and average trip length. Because the algorithm was proprietary, the pool could not independently verify whether the weighting was actuarially sound.

One of the affected drivers, Carlos M., a Houston resident who had been using ride-share platforms for three years, told a consumer advocacy group that his score dropped from 85 to 62 even though his driving habits had not changed. He had simply started driving more ride-share trips during late-night hours. The insurer's model treated that as a riskier profile, but the driver argued that his personal driving—the only driving the pool's rating plan was designed to assess—had not deteriorated. The pool's board debated whether to allow mid-term repricing at all, but the insurer argued that its contract with the drivers permitted it. The contract language, which the drivers had accepted when enrolling in the telematics program, allowed the insurer to adjust premiums based on telematics data collected at any time.

By mid-2024, the Texas Department of Insurance had received a formal complaint from one of the drivers. The regulator opened an inquiry but quickly ran into a wall: the insurer refused to share the algorithm's code or weighting formulas, citing trade secret protections. The department's analysts could examine only the loss-ratio data for the pool and the aggregate telematics scores, not the model inputs. After six months, the inquiry closed without a finding of wrongdoing, but the regulator issued a non-binding recommendation that insurers disclose telematics score components to policyholders. The insurer did not adopt the recommendation.

Why a Telematics Score Is Not a Driving Score

Telematics scores in auto insurance typically combine several factors: mileage, time of day, hard braking events, rapid acceleration, and cornering behavior. Insurers use these to estimate claim risk and offer discounts to low-risk drivers. But the score is not a pure measure of driving skill; it is a risk proxy. When a driver uses the same vehicle for both personal trips and ride-share work, the score becomes a blend of two distinct risk environments. Ride-share trips tend to occur during higher-risk hours—late nights, weekends—and involve more stops, more urban driving, and more passenger-related distractions. An insurer that blends these data points without separating trip types may produce a score that penalizes the driver for behaviors that are inherent to the job, not to personal driving choices.

Actuaries within the Texas pool raised this concern during the repricing incident. They noted that the pool's rating plan had been built on the assumption that telematics scores reflected only driver-controlled behavior. The insurer's algorithm, by contrast, implicitly assigned a risk premium to ride-share trips that the pool had not approved. The disagreement highlighted a fundamental question: should a telematics score measure the driver's behavior, or the combined risk of the driver and the vehicle's use case? The insurer argued that the score should reflect the total risk presented by the vehicle, regardless of trip purpose. The pool's actuaries countered that the pool's rates were designed for personal auto only, and that ride-share risk should be priced separately, perhaps through a commercial rider.

Some industry observers have pointed to similar cases in other states. In California, a 2022 regulatory guidance warned that telematics programs using blended scores could constitute unfair discrimination if they did not account for trip purpose. But Texas has no such guidance. The insurer in the Texas case noted that its algorithm had been reviewed by an independent actuarial firm, but the pool was not shown the review. The lack of transparency left the drivers in a bind: they could not challenge the score because they could not see how it was calculated, and the pool could not challenge it because the algorithm was proprietary.

The broader implication is that telematics scores, as currently designed, may not be suitable for vehicles used in both personal and commercial contexts. As ride-share and delivery gigs grow, more drivers fall into this dual-use category. Insurers that treat all trips equally risk mispricing risk, while drivers face premium volatility they cannot control. Some consumer advocates have called for a standard telematics score format that separates trip types, but no industry consensus has emerged.

Embedded Coverage That Triggered Without Notice

The ride-share platform used by the three drivers offered its own embedded coverage—a per-trip liability policy that activated when the driver was en route to pick up a passenger or during a trip. This coverage is common among major ride-share platforms and is intended to fill the gap between personal auto insurance and commercial coverage. But the embedded coverage in this case did not protect the drivers from the mid-term repricing of their personal policies. The platform's coverage was triggered only during ride-share trips; it did not affect the telematics score or the premium on the driver's personal policy.

The policy language in the drivers' personal auto contracts allowed the insurer to reprice based on telematics data collected at any time, including during ride-share trips. The drivers had not read that clause, and the platform had not warned them that using the app could affect their personal insurance rates. Elena R. told a reporter that she thought the embedded coverage meant her personal policy would not be affected by ride-share work. In fact, the opposite was true: the embedded coverage protected her during trips, but the data from those trips was used to raise her personal premium.

The disconnect between embedded coverage and personal policy pricing is not unique to this case. As embedded insurance products proliferate—coverage bundled into ride-share, delivery, or rental transactions—consumers may assume that the embedded policy replaces or shelters their personal coverage. But embedded policies are typically narrow, covering only specific perils or time windows. The underlying personal policy remains the primary contract, and its terms, including telematics repricing clauses, continue to apply. Regulators in some states have begun to examine whether such clauses should require explicit opt-in consent, but no rule has been adopted in Texas.

The Texas pool incident also revealed that the insurer's telematics app was integrated with the ride-share platform's API, allowing the insurer to distinguish ride-share trips from personal trips. That integration meant the insurer could, in theory, have scored the two trip types separately and applied different rating factors. It chose not to, likely because the pool's rating structure did not require it. The decision to blend scores was a business choice, not a technical limitation. The drivers, however, had no way to know that such a choice had been made.

How the Algorithm Became a Black Box for Regulators

When the Texas Department of Insurance (TDI) opened its inquiry into the repricing, its analysts requested the insurer's telematics model documentation. The insurer provided a high-level description of the variables used—mileage, time of day, braking events—but refused to disclose the weighting factors or the algorithm's code, citing trade secret protection. TDI's statutory authority to examine model inputs is limited. Under Texas law, insurers must file rates and rating plans, but they are not required to disclose proprietary algorithms that produce those rates. The department can review the output—the rates charged—but not the internal logic of the model.

The insurer defended its position by noting that the pool's rate filing had been approved by TDI, and that the telematics program was a discount program, not a rating factor. The repricing, the insurer argued, was a discount adjustment, not a rate change. That semantic distinction mattered because rate changes require prior approval, while discount adjustments typically do not. The pool's actuaries disagreed, arguing that a 40% premium increase was effectively a rate change, but the regulator did not challenge the insurer's characterization.

This regulatory gap is not confined to Texas. As of 2024, no U.S. state has mandated that insurers disclose the full structure of telematics models used in underwriting or rating. The National Association of Insurance Commissioners has issued best-practice guidance on model governance, but compliance is voluntary. Some states, like California and Massachusetts, have stricter rate review processes, but even they rely on aggregate loss-ratio data rather than model-level audits. The result is that insurers can use algorithms that produce outcomes the regulator cannot fully explain.

The Texas case also highlighted the challenge of auditing telematics models in a pool context. The pool's own actuaries had no access to the algorithm, so they could not assess whether the repricing was actuarially sound. The pool's board considered hiring an independent actuarial firm to reverse-engineer the scores, but the cost was prohibitive for a three-driver dispute. The drivers, meanwhile, had no standing to demand the algorithm under Texas law. The black box remained sealed.

The Cat-and-Mouse of Gaming the Score

When the three drivers learned that their ride-share trips were raising their telematics scores, two of them—Elena R. and David K.—attempted to game the system. They turned off the telematics app during personal trips, hoping to artificially lower their scores by showing only ride-share data. The insurer's system detected the pattern: the app was active only during ride-share trips and disconnected during personal trips. The insurer interpreted this as a fraud indicator—a deliberate attempt to manipulate the score—and flagged the drivers for premium leakage. Both drivers received non-renewal notices from the pool, effectively losing access to the shared market.

The insurer's special investigations unit (SIU) reviewed the cases and concluded that the app-disconnect pattern constituted material misrepresentation. The drivers argued that they had turned off the app to prevent their personal driving from being penalized, not to defraud. But the policy's telematics agreement required continuous app operation; turning it off violated the terms. The pool, which had no role in the telematics program, was forced to accept the non-renewals because the insurer was the servicing carrier. The drivers ended up in the surplus lines market, paying roughly double their previous premiums.

Telematics data integrity is a growing concern for the insurance industry. As more policies rely on smartphone-based telematics, the risk of data manipulation increases. Drivers can turn off the app, use a second phone, or drive without the phone present. Insurers have responded with heuristic detection—looking for gaps in data, sudden changes in driving patterns, or inconsistencies with GPS logs. But these heuristics are not foolproof, and they can penalize drivers who have legitimate reasons for disconnecting, such as battery conservation or privacy concerns.

The Texas case also showed that gaming detection can create perverse incentives. Drivers who try to game the score may be identified as fraud risks, while drivers who simply accept the repricing pay higher premiums. The insurer's SIU acknowledged that the two non-renewed drivers had not committed traditional fraud—they had not filed false claims—but argued that the app-disconnect violated the contract. The pool's board discussed whether the insurer's detection methods were too aggressive, but no changes were made. The incident underscores the need for clear rules about what constitutes telematics fraud and how insurers should handle suspected gaming.

Open Questions for Telematics Pricing

The Texas pool's experience offers several lessons for the broader insurance industry. First, single-pool pricing cannot easily accommodate dual-use vehicles. When a driver uses the same car for personal errands and ride-share work, the risk profile changes by trip. Pool rating plans that assume a single risk profile will misprice coverage. The solution may be to create separate rating tiers for ride-share drivers, with distinct telematics score thresholds and premium ranges. Some insurers have begun offering hybrid policies that automatically switch between personal and commercial coverage, but these are not yet widely available in shared markets.

Second, embedded coverage must be decoupled from personal policy pricing. The ride-share platform's per-trip coverage in this case did not shield drivers from personal rate increases. If embedded insurance is to serve its purpose, it should either replace the personal policy during ride-share trips or guarantee that ride-share data will not affect personal scores. Regulators could require that telematics programs offer a separate score for ride-share trips, or that insurers obtain explicit consent before using ride-share data for personal rating.

Third, regulators need model audit rights to ensure fairness. The Texas regulator's inability to examine the insurer's algorithm left a gap that allowed a 40% premium increase without transparent justification. States should consider requiring insurers to file telematics model documentation, including variable weights and validation results, as part of rate filings. The NAIC's model governance framework could be adapted to mandate that proprietary algorithms used in rating be subject to independent actuarial review, with results shared with regulators under confidentiality agreements.

Finally, the incident shows that the shift to usage-based insurance requires guardrails. Telematics can improve risk selection and reward safe drivers, but it also concentrates power in the hands of insurers who control the algorithms. Without transparency, consumers cannot verify that their scores are accurate, and regulators cannot ensure that rates are not unfairly discriminatory. The Texas pool's three drivers paid $2,700 for that lesson. Will regulators act before thousands more pay the same price?

This article is presented as a case study of a specific insurance pool incident and does not constitute personalized insurance or legal advice. Policyholders should consult with a licensed insurance professional or attorney regarding their specific circumstances.