Telematics Motor Pricing Under Inflation and Competitive Pressure

Conclusion: the 6% increase is not obviously sufficient and should not be signed off as a single portfolio-wide number.

• Rate need versus inflation. The current loss ratio is 78% on £14m, i.e. about £10.9m of claims. If claim severity trends at a weighted 8–9% (parts 11%, labour 7%, plus other heads), a 6% renewal increase applied to only the renewing 85% of business earns roughly 5.1% on the whole account next year. That leaves a positive rate deficiency even before any strengthening for the 18-month data period. The proposed increase may therefore protect volume rather than margin.
• Model evidence. An AUC of 0.68 is moderate, not strong. Eighteen months includes COVID-19 restrictions and a structural inflation break; the model has not seen a full cycle, so its coefficients and relativities may not be stable. Telematics data also suffers selection bias: only policyholders who accept the device are in the sample, and young drivers' exposure changes quickly.
• Competitive position and adverse selection. A 6% increase while two competitors cut 4% implies a 10-point relative price movement. The distribution team's forecast of 15% renewal loss concentrated in the best risks is the critical issue: if the best risks leave, the remaining portfolio has a higher loss ratio, so the achieved average rate increase may not improve the combined ratio. A uniform increase worsens this.
• Judgement. I would not certify 6% as adequate. I would support a segmented increase: larger increases where telematics and claims experience justify them, capped or delayed for the most price-sensitive preferred segments, with a monitoring trigger. The board should be told the central rate need is a range, not a point.

Recommended 12-month plan

1. Rebuild the pure premium on a frequency-severity basis. Model claim frequency and average severity separately, with explicit severity trend by head of damage (parts, labour, credit hire, bodily injury). Use the telematics variables only where they are stable and additive; test interactions with age, vehicle group and region. Keep a simpler GLM benchmark alongside the gradient-boosted model so the board can see whether the extra complexity is earning its place.
2. Apply a segmented rate change, not a flat 6%. Start from the indicated technical rate need by segment. For segments with high harsh-braking or night-time exposure and poor loss experience, file increases above 6%; for the best risks, hold increases at or below the market movement. Use a capped increase and a new-business discount only with a clear payback test, because deep discounts create adverse selection.
3. Add an explicit inflation load and a model-risk load. The inflation load should be scenario-based, not just a central trend. The model-risk load should reflect the 18-month data window and the AUC; a reasonable approach is to price to a range and hold the difference as capital rather than bury it in an unvalidated coefficient.
4. Validate and monitor. Back-test on a holdout period, monitor renewal retention by risk decile, compare modelled and actual claims monthly, track telematics data drift, and set triggers: if retention among preferred risks falls below 80% or severity runs more than 2 points above trend for two quarters, re-rate.
5. Collect more data and external benchmarks. Extend the experience period, obtain industry claims inflation indices, and consider a small external data purchase for young-driver severity.

The trade-off is clear: a lower headline increase protects volume but risks under-pricing; a higher flat increase protects margin but accelerates adverse selection. Segmentation and monitoring are the compromise.

Key messages for the board

• The 6% figure is a point estimate from a model with limited evidence, not a precise answer. The honest presentation is a range: for example, a central technical indication of 7–9% with a wider plausible range once model and inflation uncertainty are included.
• The commercial trade-off is material. A 6% increase is likely to lose around 15% of renewals, and the losses are expected to be concentrated in the best risks. That changes the mix and can make the remaining book worse, so the board should not assume that a 6% increase buys 6% of margin.
• I recommend a staged decision: approve a 4–5% base renewal increase plus targeted loadings, with delegated authority to move within a pre-agreed range if monthly monitoring shows retention or severity outside plan.
• The board should also understand what would change the answer: a sustained fall in parts inflation, better-than-expected retention in preferred segments, or a validated model on a longer window would support a lower increase; a further inflation spike or adverse court ruling would require more.
• I would document the advice, the range, the assumptions and the monitoring triggers, and make clear that I am not certifying the 6% as sufficient on current evidence.