Rebuilding Claims Forecasting Accuracy for a UK motor insurer in a volatile cost environment

The Challenge

UK motor insurance is a highly competitive and structurally evolving market. Price comparison websites have commoditized distribution, margins remain under pressure, and long-term outperformance increasingly depends on underwriting discipline and claims insight rather than pricing alone - particularly given that claims costs represent most of the cost base.

At the core of this is claims forecasting. Insurers that can more accurately anticipate movements in claims frequency and severity are better positioned to price risk, reserve capital and respond to market inflection points ahead of competitors.

A leading UK personal lines motor insurer engaged L.E.K. to move beyond conventional, largely backward-looking forecast approaches and build a more forward-looking, data-driven capability. The existing process lacked the granularity required to isolate underlying drivers of claims movements, limiting the ability to distinguish structural trends from short-term noise and act proactively.

Our Approach

L.E.K. delivered a comprehensive transformation of the client’s claims forecasting capability, combining deep insurance domain expertise with advanced analytical methodologies and an investor-oriented perspective on value creation.

Redesigned claims segmentation to reflect true economic drivers of frequency and severity, moving beyond legacy administrative categories
Decomposed frequency into incident rate and propensity to claim, and segmented severity at granular Head of Damage (HoD) level
Developed multi-dimensional segmentation frameworks based on road type, incident type and speed/severity dynamics, supported by alternative methodologies (e.g., severity-based and FNOL text-based classification)
Incorporated mix-shift modelling to capture the interaction between frequency movements and severity outcomes, improving attribution of cost drivers

Designed a structured, end-to-end driver-based forecasting process, spanning data capture, modelling, judgement, monitoring and feedback loops
Integrated internal claims data with external datasets including traffic flow, car parc evolution, weather, crime statistics and CPI-based cost indices
Developed quantitative models linking key claims variables to frequency and severity, supported by regression-based analysis, time-series techniques and feature engineering
Established a forecast triage framework to determine where full quantitative modelling was appropriate versus where structured qualitative judgement was required

Developed a three-dimensional PoC prioritization framework based on indemnity spend, underlying volatility and feasibility of forecast
Identified priority segments including high-speed incidents, theft and low-speed parking claims, and defined targeted modelling approaches for each
Outlined a phased roadmap covering data sourcing, feature engineering and model development, alongside enhancements to claims data capture (e.g., road type, location data) to improve long-term forecasting accuracy
Defined pathways to embed outputs into pricing, reserving and capital management processes

The Results

The engagement delivered a comprehensive blueprint for transforming claims forecasting into a forward-looking strategic capability. Key outcomes included:

A more economically meaningful claims segmentation architecture
A best-practice forecasting operating model integrating internal and external data
A prioritized roadmap for predictive model development
Clear integration into pricing, reserving and capital planning processes

By strengthening its ability to anticipate shifts in frequency and severity, the insurer enhanced forecast accuracy, reduced earnings volatility risk and improved decision-making discipline in a structurally evolving market.

Through a combination of deep sector expertise and advanced analytics, L.E.K. helped position the client for long-term performance in an increasingly disrupted insurance landscape.