Before the Renewal Notice

Worth remembering“CSAA Insurance Group has already built a simulated population of thousands of households to explore value, retention and the consequences of insurance decisions. The next question is whether the same logic can help insurers understand a price change before a real policyholder receives it.”

Somewhere inside a computer model built by CSAA Insurance Group, 5,000 households were living compressed versions of their financial lives.

They had income, cash, homes, vehicles, insurance losses, premiums and different levels of persistence. Some were single. Some were couples with children. Some retained their policies; others left. The households were not real customers, and CSAA explicitly stated that the presentation data and behaviours were notional. But the simulation itself was real: an agent-based digital twin created by a named insurer to understand value to the customer, value to the insurer and the side effects of decisions before taking real action.

At the end of its 2022 presentation, CSAA's team listed the questions it wanted the model to explore next. One stands out today: when do price changes affect value, and is the impact immediate or prolonged?

That question changes premium setting from a calculation into a rehearsal.

Insurers already know how to model risk. Many also model price elasticity and churn. What remains harder is seeing the whole customer decision: the increase, the explanation, the household circumstances, the perceived fairness, the alternatives offered and the memory the interaction leaves behind.

A premium increase has two meanings

Inside an insurer, the renewal price may be the output of a rational chain. Claims severity has risen. Reinsurance costs have moved. Repair costs, weather exposure or medical inflation have changed. The pricing team has thousands of policies to rebalance while protecting solvency and portfolio performance.

To the policyholder, the same change arrives as a number on a page.

Imagine a household opening its renewal notice and finding an 18% increase. The insurer sees a revised risk price. The customer may see a loyalty penalty, a judgement about their neighbourhood, a cost they cannot absorb or a reason to question the relationship. A carefully engineered premium can still become a badly experienced decision.

This is why price elasticity alone is incomplete. It can estimate who may leave. It does not necessarily explain what the increase means to them, which explanation they will believe, what alternative they would accept or whether a short-term retention intervention creates longer-term distrust.

What CSAA actually built

CSAA's model offers a useful non-technical picture of how this type of system works.

Instead of one average customer, the simulation represented thousands of household agents. Each household carried a combination of characteristics and could move through different states over time. The model included auto and home policies, life stage, customer segment, age, state, available cash, income, insurance losses, premium and retention.

The simulation connected those households to a decision model. External factors such as competition, the economy, regulation and the environment could influence results. Business decisions could then be assessed against measures such as profitability, cycle time, customer satisfaction and defects. CSAA combined statistical methods, machine learning and agent-based modelling to examine how actions created value and side effects.

Think of it less like asking a chatbot what customers think and more like building a small, controlled insurance economy. The insurer can change one condition, run the world forward and compare what happens across households and over time.

Modern synthetic decision intelligence extends this idea. Richer customer populations can incorporate behavioural segments, interviews, complaints, call transcripts, renewal history and customer language. Teams can expose those populations not only to a price, but also to the message, channel, timing and alternatives surrounding it.

What could be rehearsed before renewal

• The shape of the increase. Compare a single increase with phased changes, deductible adjustments or different coverage choices.
• The explanation. Explore whether customers understand loss inflation, local risk, coverage changes or market conditions - and where the explanation feels evasive.
• The timing and channel. Test whether advance notice, email, post, app messaging or an adviser conversation changes comprehension and response.
• The alternatives. Examine how payment plans, excess changes, coverage options or proactive reviews affect perceived control.
• Segment differences. Identify where long-tenured, claim-free, vulnerable, multi-policy or price-sensitive customers interpret the same decision differently.
• Second-order effects. Look beyond immediate renewal to complaints, shopping, reduced coverage, loss of bundling, future claim confidence and trust.

The surrounding evidence is already moving

CSAA is not the only signal that insurance pricing is becoming more behavioural.

Pillow implemented machine learning across its motor-insurance renewal process to predict churn and customer price elasticity, helping determine an optimal renewal price while balancing profitability and retention. That is not synthetic-customer rehearsal, but it shows that insurers are already treating renewal as an individual behavioural decision rather than a portfolio-wide arithmetic exercise.

Communication matters too. Datos Insights reported that customers who completely understood why their rate increased were 14 percentage points less likely to shop for another policy and 21 percentage points more likely to strongly agree that their insurer acts in customers' interests.

The downside of poor execution is equally visible. In 2025, Australia's securities regulator alleged that RACQ Insurance had sent more than 570,000 renewal notices containing misleading previous-premium comparisons. RACQ acknowledged the proceedings and said it had self-reported the matter; the claims remained allegations at the time of reporting.

Taken together, the signals form a clear progression: insurers can simulate household value, predict price response and observe the importance of explanation. The missing layer is bringing those elements together before the notice is issued.

How the renewal decision changes

The conventional workflow is familiar: actuarial and pricing models produce a premium, commercial teams review retention risk, legal and communications teams shape the notice, and reality delivers the answer through renewals, calls, complaints and switching.

A rehearsal model introduces another stage. Teams create several pricing-and-communication scenarios, expose them to relevant synthetic populations, compare reactions, identify fairness and comprehension risks, eliminate weak approaches and measure where confidence is strong or weak.

Where a simulation has been benchmarked and validated for a defined renewal decision, it can replace portions of conventional customer research or live experimentation.

The value is not pretending uncertainty has disappeared. It is making uncertainty more visible before the customer pays the price for it.

The first benefit is a better notice. The larger benefit is a better memory.

In the short term, the business benefits are practical: more options explored, fewer obviously weak messages, earlier visibility into segment differences, better retention interventions and faster alignment between pricing, actuarial, customer, legal, insights and service teams.

The longer-term advantage comes from connecting every rehearsal to what actually happened.

A simulation predicts which customers will understand, object, switch or seek help. The renewal cycle produces real outcomes. When those outcomes recalibrate the model, the insurer begins accumulating a proprietary history of pricing decisions, predicted reactions, actual behavior and model error.

Over several cycles, that can become decision infrastructure. Competitors may buy access to similar AI models. They cannot instantly acquire another insurer's calibrated understanding of how its own customers respond to price, explanation and circumstance.

That is the credible form of FOMO. The advantage is not simply adopting synthetic customers before everyone else. It is beginning the validation loop earlier. By the time simulation becomes standard, one insurer may be deciding with years of accumulated evidence while another is still learning when the model can be trusted.

How can Fluexy solve this

Fluexy is building synthetic decision intelligence around this principle of decision rehearsal. Teams can create relevant customer populations, expose them to competing premium, message and renewal scenarios, compare reactions and identify which decisions can be made synthetically versus which still need a real-world check.

The ambition is to replace increasing portions of conventional customer research, A/B testing, live experimentation and expert-only judgement as reliability is demonstrated for specific tasks and populations.

That last condition matters. The system should not earn trust because its answers sound human. It should earn trust because its predictions are benchmarked, its limits are visible and actual outcomes continually improve the next decision.

Have a renewal decision worth rehearsing?

If your team is preparing a premium change, renewal communication or retention strategy, bring one unresolved example to a discovery conversation.

The discussion can map the customer populations involved, the pricing and communication scenarios worth exploring, the evidence required to trust the simulation, which conventional inputs could potentially be replaced and where real-world validation still matters. No sales presentation is necessary; the decision itself is enough to start.