Personalization: Why Banking Leads Insurance & Wealth

Banking and fintech provide high-frequency, data-rich interactions that support ML-driven personalization at scale. Insurance interactions are…

May soound rube, but the honest answer is not that banks care more about their customers. It is that a bank sees you dozens of times a month while your insurer hears from you roughly once a year, and machine learning is hungry for exactly that difference.

Personalization is more mature in banking and fintech because those businesses generate the one ingredient machine learning cannot do without: frequent, structured, first-party data. A current account throws off dozens of transactions a week, each one a labelled clue about how a customer actually lives. Insurance is low-frequency and event-driven, so the data arrives in a trickle and often at the worst possible moment, a claim. Wealth sits in between: rich in data, but wrapped in advisory judgement, suitability rules and a human relationship that resists automation. The maturity gap, then, is mostly a data-and-frequency gap, not a gap in intent.

The core tension, in one line:

Personalization does not follow the firms that want it most. It follows the data, and the data follows frequency.

Use Case: A neobank can watch a salary land, notice a foreign-currency fee three days later, and offer a travel card before the customer has finished booking the trip. A motor insurer typically has two scheduled data moments a year, the renewal and, if things go wrong, a claim. Same appetite for personalization; completely different raw material to work with.

1. The real driver: interaction frequency and data richness

The clearest way to see the gap is to line up the four contexts by how often the customer shows up and what that contact leaves behind.

Sector	How often the customer interacts	Data it generates	Personalization ceiling today
Banking & fintech	Daily to weekly	High-volume, structured, first-party transaction data	High: real-time, ML-driven offers and nudges
Insurance	Once or twice a year, plus claims	Sparse, event-driven, split across underwriting and claims	Low to moderate: mostly at quote and renewal
Wealth	Periodic reviews, plus market-driven contact	Rich but complex: portfolio, suitability, planning, held-away assets	Moderate: capped by advice rules and the adviser relationship

This is why personalization matured first where the data was densest. According to McKinsey, 71 per cent of consumers now expect personalized interactions and 76 per cent are frustrated when they do not get them. Banking and fintech could meet that expectation because every swipe, transfer and login is a fresh, labelled training example arriving in real time.

2. Why insurance lags, and it is not for want of trying

The data arrives too rarely. A once-a-year touchpoint gives a model almost nothing to learn from between renewals, so a model has little to act on.
It is trapped in silos. Underwriting, policy administration and claims often sit on separate systems, so no single, current view of the customer exists to personalize against.
The moments are emotionally charged. Much of an insurer’s contact happens at a claim, where personalization can feel intrusive if it is handled clumsily. EY notes there is a thin line between hyper-personal and over-personal, and cites Gartner research that 38 per cent of customers will stop dealing with a brand if its use of their data feels uncomfortable.

Example: usage-based motor insurance is the exception that proves the rule. Once a telematics app streams daily driving data, the insurer suddenly has banking-like frequency, and tailoring accelerates in step.

3. Why wealth is a different problem from insurance

It is tempting to lump the two laggards together, but wealth is not data-poor. It is judgement-heavy and rule-bound, which is a different constraint entirely.

The data is rich but fragmented. Portfolios, suitability records, financial plans and held-away assets rarely live in one place, so the blocker is integration, not collection.
Suitability and advice rules cap automation. A recommendation is a regulated act. You cannot let a model nudge a client into a product the way a retailer nudges a shopping basket.
The relationship is the product. Clients pay for a trusted human, so personalization has to augment the adviser, not appear to replace them. Deloitte makes a similar point across financial services: scaling from a promising pilot to a personalized experience that every stakeholder trusts is the real hurdle, not the model itself.

4. A contrarian note: “mature” is not the same as “better”

The league table flatters banking, and it is worth pausing on why.

Frequency breeds shallow personalization. A great deal of bank personalization is next-best-offer nudging. It is mature and well-industrialised, but it is not always deep or especially loved.
Scarcity can force depth. Because wealth and insurance touch the customer rarely, each interaction carries more weight. A genuinely tailored annual review, or a claim handled as if the insurer understood your life, can build more loyalty than a hundred app nudges.
The gap is closing from both ends. Embedded insurance, open banking and open finance data sharing, and connected devices are quietly handing insurers and wealth managers the frequency and the data they always lacked. The head start is real, but it is not permanent.

5. What actually closes the gap

None of this is a counsel of despair for insurers and wealth managers. It is a different order of operations.

Fix data integration before buying an AI engine. In insurance and wealth the binding constraint is a unified, current customer view. A recommendation engine sitting on top of siloed data personalizes badly, confidently.
Manufacture the missing frequency. Telematics, connected-home data, open-banking feeds and portfolio aggregation create the regular interactions these sectors were never structured to produce.
Personalize the high-stakes moment, not just the offer. A claim or an annual review is precisely where insurance and wealth can out-personalize a bank, if the data and the design are ready.
Keep the human in the loop where advice is regulated. Position models as decision support for advisers and underwriters, with suitability and an audit trail built in from the start.
Respect the creepiness line. Consent, transparency and restraint matter more as the data gets more sensitive. Over-personalization destroys trust fastest in exactly the sectors that can least afford to lose it.

Personalization did not mature fastest where firms cared most. It matured where customers showed up most often. The task for insurers and wealth managers is not to copy the bank’s nudges, but to manufacture the frequency and unify the data they never had.

The maturity gap is real, but it was never really about ambition, and it is closing. Banking and fintech had a head start because their business model produces the dense, first-party data that machine learning turns into relevant, timely experiences. Insurance and wealth begin from sparser or more fragmented data and stricter rules, so their route to it runs through data integration, new data sources and careful design, not through imitating retail-style nudging.

This is the kind of problem Optimizium is built for: unifying fragmented policy, claims, portfolio and planning data into a single current view of the customer; adding the data feeds, from telematics to open finance to portfolio aggregation, that create the frequency these sectors lack; and designing personalization that respects suitability, advice rules and the trust these relationships depend on. The aim is not personalization for its own sake, but the right message, to the right client, at the moment that matters, done in a way that both a regulator and a customer would be comfortable with.

Explore how we approach this across Banking, Insurance, Wealth Management and Beyond Banking, browse the FAQ Hub, or see the related question: What is the role of portfolio platforms in wealth digitalization?