Quantee has explored two distinct approaches to performing price optimisation in insurance, focusing on methods that avoid the controversial practice of price walking while still adhering to business objectives for conversion and retention. This article builds on the ethical considerations of price optimisation and compares the benefits and trade-offs of separate versus joint optimisation models.
Price optimisation involves adjusting profit margins across an insurance portfolio to maximise overall profit, within legal and strategic constraints. However, this approach has long raised ethical concerns, particularly when it results in some customers paying more for the same coverage based solely on their perceived willingness to accept higher prices.
Less price-sensitive clients might not question a higher quote due to trust in the insurer, limited access to alternatives, or simple financial ease. Meanwhile, more price-sensitive customers invest time into seeking better deals.
This naturally leads to ethical dilemmas—should insurers charge trusting or loyal customers more simply because they can?
One widely criticised practice derived from this logic is price walking, where renewing customers are charged more than new customers.
While this may yield short-term profits, it can harm brand reputation and customer loyalty. It’s no surprise, then, that price walking has been banned in markets like the UK.
Avoiding price walking in optimisation
The goal is to perform price optimisation that does not discriminate based on whether the customer is new or renewing, while still allowing insurers to set and meet sales targets. To do this, constraints on individual prices and overall demand are applied. While new and existing customers often exhibit different demand behaviours, the optimisation algorithm must be blind to their status.
This presents a challenge, since new and existing customers also differ in their characteristics. This requires careful modelling so that demand is controlled without biasing prices based on customer tenure.
Comparing two optimisation strategies
Quantee proposes two approaches to tackle this problem:
-
Separate optimisation – New business and renewals are treated independently. Each dataset is modelled and optimised separately based on their own constraints. After optimisation, the two are merged, and a regression model is used to calculate a midpoint price for each profile, effectively averaging the two price points based on likelihood of conversion or renewal.
-
Joint optimisation – Datasets for new and renewing customers are merged but retain separate purchase indicators. Conversion and retention models are trained on this combined data. The algorithm is then optimised with constraints applied to total conversions and retentions. This setup allows insurers to test different demand scenarios and trace the efficient frontier of new versus renewal business performance.
Both strategies avoid assigning final prices based on whether a customer is new or returning. Instead, they manage business volume targets while maintaining a uniform pricing process.
Performance comparisons
In simulations using identical actuarial cost models, joint optimisation proved better at meeting both conversion and retention constraints, though it sacrificed some margin on new business in favour of better performance with existing clients. Separate optimisation gave the highest margin when scored on the dataset it was specifically trained on, but fared worse when applied across portfolios.
Moreover, the joint model supports scenario testing and provides more consistent constraint management. However, it is computationally more complex and sensitive to demand model inconsistencies. The separate optimisation method is more straightforward, relying on a simpler regression model at the final pricing stage, making it easier to explain and quicker to implement.
Final considerations
While neither method involves price walking, Quantee cautions that price disparities may still emerge in aggregate—certain market segments might systematically receive higher or lower margins based on the demographic skew of new or existing customers. As such, careful modelling is essential.
Furthermore, results can vary based on how demand models are structured. More accuracy might be gained by decoupling the estimation of customer status from their pricing sensitivity, training models separately on conversions and renewals, and then combining them to estimate the final acceptance probability.
By refining these techniques, insurers can ethically meet their business goals—retaining loyalty, avoiding regulatory pitfalls, and optimising profitability without exploiting customer trust.
Read the full blog from Quantee here.
Copyright © 2025 FinTech Global
