The hidden costs of actuarial modelling systems

Why understanding total cost of ownership means thinking beyond software licences to people, processes and risk, as Joshua Geer reports

In an era where data-driven decisions underpin every move in insurance, actuarial modelling systems have never been more vital or more complex.

But while these systems help insurers assess, price and manage risk, the cost of operating them is often only partially understood.

It's tempting to see these systems as self-contained or simply a line item in a software budget, but experts say this view misses significant layers of cost – not just technological, but operational, regulatory and human.

Understanding these hidden dimensions is essential for insurers that want to remain competitive, scalable and compliant in a fast-evolving market.

InsuranceERM has spoken to actuaries spanning the market to explore how insurers are defining and managing the real cost of actuarial modelling, where they're still falling short, and how future developments could reshape the cost landscape.

What does total cost of ownership really mean?

The first challenge is one of scope. While the first focus will inevitably be on licensing, infrastructure and vendor fees, when thinking about total cost of ownership, this view is dangerously narrow.

"You have actuarial systems in that 'little box' of your IT infrastructure, but there's a lot of other IT infrastructure that underlies it," says Ronald Richman, former chief actuary of Old Mutual and founder of consultancy InsureAI.

He relates an experience of moving a data system from one platform to another, which broke the inputs to the actuarial system. "So the total cost of ownership needs to be end-to-end in the value chain. It can't just be 'I'm paying for this little box' and ignoring the technical debt that can be accumulating in your data flows, in your reconciliations, in your financial systems that actually enable good actuarial work to be done," he explains.

This end-to-end view is echoed by actuary Cecilia Wang, who is head of pricing and longevity at French insurer Scor.

For her, the costs of actuarial systems span four major categories: "First is IT cost, things like software licences, IT infrastructure and IT support. Second is system development and maintenance, which is a big chunk of the cost. Third is operational cost, the day-to-day use of the system. A more efficient system reduces this cost. The fourth area is governance and regulatory costs (audit, approvals, controls) which are a big part of the agenda for large companies."

Still, these conventional categories fail to capture what Wang and others argue is one of the most significant hidden costs: humans.

For example, she highlights the impact of missed opportunities due to limited human bandwidth.

"The longevity market is so busy that pricing teams have to choose which opportunities to take. So, it's critical that our people spend time on high-value tasks like applying judgment or analysing results, not on repetitive or manual work," she says.

Jeremy Levitt, CEO of the consultancy network Graeme Group and an actuary and former director at Axa, agrees the total cost of ownership is often misunderstood or underexplored.

"The total cost of actuarial modelling systems is determined by implementation costs, ongoing licensing fees, the cost of running the system on the cloud, vendor consulting fees to resolve issues/maintain the modelling infrastructure and the opportunity cost of not using other, potentially more efficient platforms," he says.

"There are also indirect and qualitative costs, such as the flexibility of the contract terms in place with the software vendor, ease of integration, the level of support received by the vendor, and the ease of governing the model."

The silent drains

As the quotes suggest, people – and what they are or aren't enabled to do – represent one of the most substantial, yet overlooked, cost drivers.

Scor's Wang elaborates: "Actuaries are expensive, and they need to be involved in both development and maintenance. And for system development projects especially, they can easily last over one year or two years, if not longer.

"The primary cost driver for my team, is the number of opportunities we are unable to price because of human bandwidth."

This "opportunity cost" doesn't appear in traditional budget reports but has a real and material impact on business growth and deal success, says Wang. The inability to deliver fast and accurate pricing in competitive markets like longevity could mean losing business entirely.

Benchmarking

Ben Sheldon, group chief actuary at speciality re/insurer Convex added while direct costs such as software licensing, implementation and maintenance are "relatively simple to assess," he noted understanding the significant cost of staffing is complex.

"Insurers will assess how many actuarial professionals are required to run their processes and often compare team size relative to the scale of their reserves or operations as a benchmark against peers," he said.

But Sheldon explained benchmarking in this fashion can be imprecise, as the workload required to run a model varies based on factors such as the number of legal entities, lines of business and responsibilities split between the actuarial and finance teams.

"In practice, it's a difficult and resource-intensive process to compare the costs of modelling systems and their demands. While insurers sometimes undertake this process during major overhauls or RFP processes, doing it on a continuous basis is unrealistic due to its complexity. Ultimately, most insurers rely on peer comparisons and internal efficiency metrics, like headcount or turnaround time, to assess and manage costs," he stated.

Taha Ahmad, an experienced actuary currently working at Verisk, highlights the traditional benchmarks used to assess actuarial costs, such as licensing per actuary or gross written premium (GWP) per head, are increasingly ineffective.

"For a profession which is very numbers driven, it's not a very numbers-driven thing to do – to estimate the cost of ownership of actuaries," he says.

"Often people use things like number of actuaries or GWP per actuary, and then you benchmark that," Taha says. But he explains the expense ratio model can be very different depending on the type of insurer.

As such Ahmad adds, beyond cost metrics, it's crucial to define "value-add criteria" that reflect the actual contribution of actuarial teams. "That could be numerical, commercial, or a complementary service to the front-of-house trading role... understanding what the value drivers are and making sure you're assessing against those is essential."

Technology's promise

Despite widespread belief in technology's ability to reduce costs, many insurers have not yet seen improvements in expense ratios.

"As an industry, we haven't been great at reducing our expense ratios," Ahmad says. "They haven't really shifted a lot, even with the growth and rate increases we've seen".

Where savings do occur, they are often eroded by implementation inefficiencies, he adds.

In efforts to reduce system costs, some vendors have worked on optimising data storage and management, a strategy that can help lower licensing fees by shrinking the overall data footprint.

Ahmad notes: "One thing we've done is improve how our systems store and manage data... we've been able to reduce our licensing costs significantly, because we've reduced data storage costs significantly."

One cost that has surprised some is the cloud hosting fees. Christo Muller, a partner at MBE Consulting who has worked on many actuarial software implementations, says there is now more understanding that cloud-based systems are metered and may not produce the expected savings.

"Because most actuarial platforms are not cloud-architected and effectively are just re-hosted from on-premise data centres to cloud data centres, that's been one massive cost increase I see for a number of insurers."

An added complication is that firms often incur new costs through third-party involvement.

Implementation projects may require specialist consultants or temporary staff, which can quickly absorb any savings gained on licensing.

Ahmad says: "What ends up happening is they hire contractors or consultants, which eats into the savings they would have seen from reduced licensing costs. So yes, we're charging a lower fee for licensing. But then the client uses third parties to implement it and that cuts into the saving."

Wang agrees that technology can be a cost saver but only when tightly integrated.

At Scor, her longevity team chose to develop a proprietary pricing system with a modular architecture, shared across markets. "This structure makes things much more efficient, especially for ongoing development... if we want to make a change to a central module that applies to five countries, we can do it once rather than five times."

The system is fully end-to-end and automated, eliminating tool-switching and manual error. "That reduces human error, which is actually one of the biggest hidden costs. A mistake can lead to reputational risk or significant financial loss."

And AI is starting to play a material role, says Wang. "AI tools like ChatGPT are already very good at generating initial code for large, complex projects... That saves a lot of time," Wang notes.

Sheldon added the integration of machine learning and data allows actuarial teams to access and analyse data frequently moving away from quarterly "Big Bang" processes toward continuous analysis.

He said Convex, for example, has reduced its end-to-end quarterly reserving process to just over a week, this includes production of the reserving committee paper and management information for the wider business. "The frequent analysis, and identification, of data trends are relevant for the whole business, not just the actuarial team".

Levitt adds that in the longer term, technology should improve efficiency, but only if complexity is managed.

"Generally, we expect the impact of AI and evolving technologies to reduce the total cost of actuarial modelling systems over time. Model runs will become significantly faster to extract and model output easier to interpret," he says.

However, he cautions this may also raise the bar: "Advanced technology will enable actuaries to increase the level of sophistication of their models.

It will become easier to use first-principles methods in actuarial modelling instead of approximate or simplified methods."

Future-facing risks and rising governance pressures

Even as systems become more advanced, new layers of cost – particularly around model risk – are emerging. These are now critical concerns for both regulators and boards.

"If you have a highly automated model, there's a greater risk that actuaries use it as a black box, and they don't understand what's inside," warns Wang. "To address that, we make sure every actuary works on all areas, including model development and maintenance... That builds familiarity and reduces risk."

Sheldon also noted as new tech is adopted, governance challenges will emerge. "Faster cycles mean less time for traditional checks, so we are also working on embedding appropriate controls into automated systems". He said Convex, for example, is developing governance frameworks to support its move toward continuous reserving.

It is clear version control, end-user computing controls, and audit trails are also becoming essential.

Climate risk and regulatory stress testing are further adding cost and complexity, says Wang. "Climate risk assumptions, modelling, and stress testing are all getting more complex. This leads to more complex models and longer run times, which adds to cost."

Yet the market is moving in the opposite direction, demanding faster pricing and onboarding. Firms unable to match this pace face not just cost inefficiency, but strategic risk.

According to Convex's Sheldon, at this stage, advanced actuarial modelling should incorporate ML and AI capabilities to make the actuarial process smarter and leaner, thus keep the actuarial headcount constant while producing more analysis.

"The cost of adding ML and AI capabilities into actuarial modelling will be less than the cost savings of a smarter and more engaged actuarial team," he stated. But he emphasised, linking back to his point about ensuring governance, that future cost savings will depend not just on the technology itself, but on how effectively it is integrated, governed and scaled across the organisation.

Still, the actuaries voice optimism that innovation and competition will push costs down over time.

"I'm hopeful that as more players enter the technology space, competition will drive down prices over time," says Wang.

Ahmad adds: "When you combine licensing savings, operational efficiency, and more efficient implementation, and once all of that becomes business as usual, I think we'll really start to see a difference and systems costs come down."

Towards a more strategic and integrated view

What actuaries have made clear is that the total cost of ownership for actuarial modelling systems is not just a budgeting problem. It's a strategic challenge that touches nearly every part of the insurance enterprise from system design and staffing to risk governance and competitive execution.

"A persistent problem for actuarial modellers stems from the need to work within a traditional framework, whilst looking to innovate for the future," says Sheldon.

As insurers race to modernise, those that take a more integrated, forward-looking views of total cost of ownership recognising the interplay of technology, people, and regulation will be the ones best placed to manage today's costs and tomorrow's competitiveness.