Optimized Pricing : A Key Lever for Profitable Growth CAS Spring Meeting 18 May, 2004
P&C personal line insurers around the world are seeking to set “optimized” prices to drive profitable growth … How do we set optimized prices – ie those which best meet our financial objectives eg… • Gain market share while ensuring adequate returns? • Increase returns while maintaining adequate market share?
…. But are facing significant obstacles • Don’t have a rigorous way to estimate price elasticity of granular customer segments , and to use this in price setting • Have a reasonable handle on claims and expenses, and therefore unit margins for different segments • But don’t have a robust way of estimating the way volume and therefore profit responds to price changes by segment • And don’t have a way to integrate unit margins and elasticity to predict the financial impact over time of price changes for the many thousands of segments • Don’t feel able to set optimized and therefore differential margins across segments in highly regulated environments like the US • Regulators won’t approve prices that are “unfairly discriminatory” – ie that are significantly out of alignment with costs
While customers differ both in terms of their unit costs and their price sensitivity, most insurers model the former but not the latter ? Observed Closure or Retention Rates = Elasticity Examples of Unit Cost Differences Examples of Price Sensitivity Differences • Some drivers are more risky than others (ie have higher expected claims costs) …. • Are less careful and/or less skilled • Drive more miles • Drive cars that are more expensive to repair • On routes/locations that are more accident or theft prone • Factors typically correlated with these differences include gender, age, vehicle type/age, use, location, driving/claims history, credit history etc • Some customers are more price sensitive than others (ie have a larger change in conversion /renewal rates for a given price change) • Shop around at each renewal or at low levels of renewal price increase • Shop a wider basket of competitors • Are more inclined to switch brands for small price differences • Factors likely to be correlated with these differences include income/wealth, sum insured, age, tenure, channel, extent of product bundling, payment method, credit history etc Claims models
Quantifying these segment elasticity differences enables the setting of more optimal prices Price to Max Profit at Current Volume $115 $95 $15 $35 = $4400 profit 130 70 = 200 units Price to Max Volume at Current Profit $110 $90 $10 $30 = $4000 profit 160 80 = 240 units At Current Prices Elastic Segment A Inelastic Segment B Price $100 $100 Or Margin $20 $20 = $4000 profit Volume 100 100 = 200 units Elasticity 6 2 = % ∆ in Volume for a -1% ∆ in Price
So how do you go about setting and maintaining optimal prices? Setting Initial Optimized Prices Keeping Prices Optimal 1. Build unit profit models – contribution per customer if they accept at a given price, for each segment 3. Integrate the models into an multi-year profit simulation - to determine the optimized prices 4. Continue to update models and reset optimized prices 2. Build price elasticity models – volume of customers accepting at a given price , for each segment
1. Build unit profit models to estimate the unit contribution for each applicant segment – if they accept at a given price Profit $ Profit Contribution per applicant accepting Price Expected Claims All Other Net Costs Net Cost • Model claims cost based on past claims experience, enriched with external data eg census, perils, credit • Allocate the variable component of expenses, reinsurance, investment income and cost of capital • Model the cross-sale/ cross- “unsale” value For each set of applicant characteristics
2. Build price elasticity models to estimate the price/volume trade-offs for each applicant segment – probability of acceptance at a given price • Simple approach : use past history of price changes and “strike” rate impacts, with linear models (eg GLM) • But this doesn’t work well !!! : • Past history too sparse, uncontrolled for known competitor rate changes, and massively co-linear • Well-fitted GLM strike rate models produce inaccurate elasticity predictions • More advanced approach : • Enrich past history with price variation, competitor prices and external data • Use non-linear models Number of Applicants Accepting Quoted Price Total Available Market Renewal For each set of applicant characteristics New Business Price Competitor’s Price
3. Integrate the models into an multi-year profit simulation to determine the optimised prices Unit Profit Per Applicant Accepting Price Number of Applicants Accepting Year 1 Renewal Year 2 Year 3 New Business Price • Simple approach : use segment-average elasticities x margins for a series of one-way cuts to estimate profit and volume impact of a price change • But this doesn’t work well !!! : doesn’t allow for “adverse selection” effects where elasticity and margin are correlated within a segment; doesn’t allow for “stacking” the impacts of multiple price changes • More advanced approach : • Use a granular simulator running the models over every quote record in a large sample to estimate multi-year impact • Use a visualiser to examine the impact by aggregating into segments • Use an optimiser to identify the set of price changes that best meets the financial objectives and constraints Total Profit X New Business Renewal Price For each set of applicant characteristics
The results have been outstanding Australian Home Insurer UK Auto Insurer Wanted to adjust pricing to maximize growth while maintaining a 15%pa ROC Wanted to maximize profits without shedding too much share Objectives : Raised annual profit before tax by 3% of NWP while holding share versus a control group Has grown over 40% in last 2 years at or above 15%pa ROC versus a control group Results : In general, extra pre-tax profit from Optimized Pricing around ~2% to 6% of Premium p.a., from higher margins and/or higher volume
But how can this approach work in the US market where regulations prohibit prices that are “unfairly discriminatory” ? • Despite the regulations against “unfairly discriminatory” prices, we see high variation in margin across segments for a given insurer • This arises from different competitive conditions, and the insurer’s different position and ambitions in different segments Some “discrimination” possible
There are ways to make elasticity-based pricing decisions that are not “unfairly discriminatory” Straightforward Price Changes Where optimal price increase is > cost increase • Most price reductions (ie segments with high elasticity and sufficient margin) even in segments where costs haven’t reduced as much or at all. Regulators usually like to see at least some consumers getting lower prices, even if this somewhat expands margin differentials versus other consumers • Some price increases (ie segments with low elasticity) to those segments where costs have increased by as much or more. Regulators will generally approve preservation or reduction of margins. • If your conversion/ retention rates are high and/or your prices are low vs competitors, may be able to argue that current margins are inadequate ….and so price rises in excess of the cost increases can be justified • Failing this, the price increase may need to be limited to the cost increase - ie in the right direction, but short of “optimal”… and use marketing/service levels to shift mix to more optimally priced segments
Even if you don’t use elasticity differences to set differential margins, there are other ways you can use elasticity insights to make better decisions Use the overall value models (including elasticity) and the simulation tool …… • To explore the profit and volume impact of alternative “regulated” pricing strategies – eg will your next planned rate filing really meet your financial objectives for profit and growth? How will the mix of risks change as a result of the price change? which components of the price change are actually value destroying? • To identify the best segment allocation of your marketing spend, given your marketing response models and a “regulated” price set – ie how should you allocate marketing spend not just to get the highest response rates from low risk segments, but to get the highest overall profit contribution?