Nimbus Direct Insurance Business Plan › Section 9 › Underwriting & Actuarial Strategy

9.1 Underwriting Philosophy

Nimbus’s underwriting philosophy can be summarised in five tenets:
(i) every risk is priced individually using all observable signals
consistent with regulatory and fair-practice constraints; (ii) the
loss-cost component of the rate is derived from causal exposure-based
models rather than purely correlative regressions; (iii) the expense and
capital-cost components of the rate are transparent and recalibrated
quarterly against actual experience; (iv) the underwriting result is
monitored daily at portfolio, segment and risk-factor level; and (v)
corrective action is taken within 30 days of any material deviation from
plan.

9.2 Pricing Architecture

The pricing engine is composed of four sequential layers:

1. Frequency model — Generalised Linear Model (GLM) using
   log-Poisson link; rating factors include geographic risk grid (1 km²
   hexagons), vehicle make/model/age, driver age, gender (where
   regulator-permitted), claim history, credit-based score, and telematics
   behaviour score.
2. Severity model — GLM with log-Gamma link; rating factors include
   sum insured, vehicle parts cost index, postal-code repair-shop cost
   index, and historical severity by claim type.
3. Loading factors — expense loading (target 22% at Year 5),
   commission/aggregator loading (variable by channel), capital-cost
   loading (target 4% based on SCR consumption), and profit loading (target
   8%).
4. Renewal optimisation — for in-force policies, a churn-sensitive
   optimisation overlay subject to maximum year-on-year increase caps and
   Treating Customers Fairly principles.

9.3 Reserving

Statutory and IFRS 17 technical provisions will be calculated by
Nimbus’s in-house Actuarial Function reporting to the Head of Actuarial
Function (HAF), an independently-appointed Fellow of the Actuarial
Society of South Africa. Reserving methodology will follow standard
practice for short-tail non-life lines:

• Chain-ladder (incurred and paid) and Bornhuetter-Ferguson for
  outstanding claims provisions.
• Frequency-severity simulation for newer products with limited
  claim history.
• Risk-adjustment under IFRS 17 calibrated to a 75th percentile of
  the underlying loss distribution.
• Independent peer review of year-end provisions by an external
  actuarial firm (Ernst & Young Actuarial or Deloitte Actuarial
  preferred shortlist).

9.4 Target Underwriting Ratios

Figure 9.1

Nimbus’s target combined ratio of 79% by Year 5 is composed of a
target loss ratio of 57% and a target expense ratio of 22%. Sensitivity
to each ratio component is material to net profit, hence the disciplined
focus described above. A 1 percentage-point movement in the loss ratio
in Year 5 affects net profit by approximately ZAR 37.4 million (5.0% of
base-case net profit).

9.5 Catastrophe Risk Management

Catastrophe exposure is a material risk for any South African insurer
with property exposure. The 2022 KwaZulu-Natal floods generated industry
losses estimated at ZAR 7 billion, and 2024 hailstorm losses in Gauteng
were estimated at ZAR 1.2 billion. Nimbus’s approach combines (i)
selective geographic underwriting (no aggregation in highest-risk
flood-plain post-codes), (ii) the catastrophe excess-of-loss treaty
described in Section 5.4 providing ZAR 800 million of cover above a ZAR
30 million event retention, and (iii) full participation in the SASRIA
pool for special-risk perils. The Company will hold a stress-test
reserve calibrated to a 1-in-200-year catastrophe event under the SAM
Standard Formula, with internal model development planned for Year
4.

9.6 Fraud Detection

Insurance fraud is estimated by the SA Insurance Crime Bureau (SAICB)
to cost the industry approximately ZAR 4–6 billion annually. Nimbus’s
fraud detection layer is built on (i) rules-based screening of
suspicious claim patterns, (ii) graph-network analysis to identify
connected claimants, and (iii) machine-learning models trained on
industry-shared fraud signals. Year 2 introduction of an enhanced fraud
detection v2 platform will integrate computer-vision validation of
damage photographs and audio-stress analysis of First Notification of
Loss recordings, with expected impact of a 1.5–2.5 percentage point
loss-ratio improvement.

Confidential — this business plan is provided to prospective investors and lenders for evaluation purposes only and may not be reproduced or distributed without the written consent of Nimbus Direct Insurance Group (Pty) Ltd.