The Data Science and Analytics industry needs Actuaries

Why I joined Dataly Actuarial

20 January 2023

Albert Suryadi

Associate Director, Data

Who am I

Hi everyone! My name is Albert Suryadi.

For those of you who want to know a bit more about me, here are some key highlights:

  • I studied and qualified as an actuary;
  • I've worked in data science and analytics for more than a decade;
  • I'm based in Sydney, Australia.

My Data Science Journey

Before I studied Actuarial Science, I wanted to be an entrepreneur. I spent my early childhood in Indonesia. There, the majority aspires to run their own business. My parents have one, my neighbours have one, and I thought I needed one.

During school, I am always fascinated with business studies and economics. In addition, I was also really good at maths. I decided that blending the practices would be helpful, like how do we make quantitative-driven decisions. Then I spent my undergraduate studying actuarial science, business strategy, and economic management.

After graduation, I landed jobs in a non-traditional actuarial field. While working there, I started to realise the corporate challenges, especially when it came to incorporating new technologies and quantitative methods that would supposedly improve their revenue. I became curious about how we could bring together people and data seamlessly.

The following important chapter of my career was an official analytics role at Citibank, a major global financial institution. I would go on to work in their advanced analytics team, a centralised team on all the retail banking functions such as Credit Cards, Risk Management, Collections, Fraud, Customer, Finance, Sales, and many more.

These teams were very different regarding subject areas, size, and data/technology maturity. However, I encountered similar problems, such as data access, metrics definition, tribal knowledge, automation, data literacy, and business adoption.

Many of these challenges were magnified in my career as I progressed into an Advanced Analytics and Data Science manager. As a leader, my primary responsibilities were to build and scale data solutions to drive data-driven decisions in a sustainable and repeatable fashion.

I was now responsible for end-to-end project deliverables and answerable to many more business stakeholders. I realised that the business requirements are often vague or Hollywood-inspired, leading to significant pre-work before landing in the team's actual delivery queue.

Some of the other most significant challenges we faced included insufficient engineering support, the difficulty in existing tooling, the lack of business support/availability, and coordinating the numerous moving pieces of a data science project as a lean and new team on an island.

Eventually, I began to see my experiences and battle scars working to deliver the data science hype coming together as a career in and of itself. In my subsequent roles, I had the opportunity to build (or rebuild) data analytics and data science teams for various forward-thinking organisations. I was inspired to go beyond the technical side and dive even deeper into the world of Data Science and Analytics.

The future of Data Science and Analytics

The world of Data Science and Analytics is still maturing. Although various best practices and tools exist, industry adoption is still catching up. In particular, I'm intrigued by the following opportunities and trends:

More Governance

Since 2020, there has been a noticeable rise in cyberattacks and data breaches. This includes high-profile customer data breaches on major organisations such as Optus and Medibank in Australia.

As a result, organisations are ramping up their governance. Initially, most of the initiatives have been defensive, but it's important to have a balanced approach. The overall goal of data science and analytics is to drive business value, and a good governance initiatives must enable it.

Some maladaptive governance practices I've observed include:

  • Placing all the burden on the data team;
  • Sharing responsibilities between data and business team, but the business is not keen to invest the time or be enabled to collaborate, thereby eventually converging back to the prior pattern;
  • Or defaulting to hiring only candidates with a strong engineering background only - data is a team sport and requires individuals with commercial empathy; otherwise, it'll eventually lead data back to a swamp.
  • A possible approach is to design your governance approaches with the business. Instead of only seeing them as approvers, empower your businesses with the tolls and knowledge to collaborate. This includes further data democratization, self-service user tools, data lineage, data catalogue, and others.

Several useful frameworks have also been established by overseas regulatory bodies such as Singapore, the UK, and Canada. Guiding policies such as transparency, user-friendliness, and agile could go a long way.

Put it simply: make it easy for everyone to collaborate and do the right thing.

More Frameworks

By design, machine learning is all about rules. It's no secret that data science and analytics need data, sometimes a lot. On this data, algorithms will find patterns that output a model scoring to be encoded for further decisioning rules.

It makes sense for data science teams to adopt rules and frameworks themselves. Initially, data science and analytics were born out of innovation. We had the flexibility to do everything to get the job done. However, over the year, we have amassed knowledge on what has worked and hasn't. These frameworks help the data team to deliver powerful outcomes consistently.

It also makes sense to create frameworks for human-and-AI collaboration. AI, by itself, is not that smart. Also, decisions are not an exact science. We need to design ways so humans can explore, interpret, and adjust AI outputs. Product principles, such as human-centered design (also called HCD), guide the problem-solving methods that improve delivery quality and user happiness.

More AutoML

Do we believe that AI is smart enough to automate decisions?

What does a mature AutoML tool look like?

The common advice we give stakeholders is that "everything is possible with AI." But do we take our advice in applying AI and automation for ourselves?

Everyone who says AutoML is not real is like telling me I should walk to work every day. I prefer to use our industrial-age advancement of motorised vehicles.

So what role should a data scientist play?

For an organisation focused on deriving value from their ML, the goal is to increase data science and analytics by significantly reducing the development time with a small (and increasingly scarce) number of data scientists. In this case, AutoML works together as an automated army of data scientists that accelerates model development and operationalises the solution.

Dataly Actuarial: A Data and Actuarial Consulting Company

I'm excited about the opportunity to partner with Dataly Actuarial to address the opportunities above because:

Trust is currently (or going to be) a key focus for organisations

When I started working on data science and analytics, I had no opinion on data trust. It seems like a vague concept, and the data should be trustworthy, right?

However, I've experienced all the following pain points in delivering data science and analytics projects:

  • Explaining a data or model insight requires explaining how the data sources, metrics calculation, and security. Often this took the majority of the presentation time rather than the actual insights;
  • Sharing data points and metrics with others result in disagreement over other published numbers or the "single source of truth" while not having any documentation to understand the different logic;
  • Most importantly, trusting and remembering the data logic and model with my future self.

When data trust exists, it can:

Increase data-driven decisions by making it easier to trust and use the data created rather than never-ending scrutiny;

Make collaboration delightful, especially with other data teams, because we understand that all numbers serve a purpose, and we could compare the logic;

Preserve data governance, by leveraging data access and lineage in case the audit or compliance team needs to validate our policies.

More and more technologies

The right technologies are crucial to help us solve problems faster and easier. The rise in component-based/microservices architectures accelerate these developments further as each pieces are independent.

However, we have too many options, and teams are less confident in choosing a tool. As we've known, more options aren't necessarily better - and when we are flooded with options, we are not only less likely to switch to a new option but are more dissatisfied with our choice than if they'd been presented with a more curated selection.

The 2021 Machine Learning, AI and Data (MAD) Landscape.
Source Link:

Increasing focus to demonstrate the value

Regardless of how much we complain about data science production's difficulties, it's never been faster or easier to deploy really powerful data solutions.

Some reasonable questions to ask are: Which organisations are going to deliver and iterate the fastest? What should we focus our iterations on?

The answer is value. We have moved beyond real-time capability. Speed-to-insights takes this beyond a largely technological problem to a whole different level.

And competition over organisation funding is increasing due to constrained capital. Data Science and Analytics teams don't enjoy the luxury of time we had before. The winning teams will be the ones that demonstrate the business value the fastest and reliably.

So how does Dataly Actuarial help, rather than add, to the chaos?

  • Core focus on driving business outcomes, beyond the technology and numbers: in a complex environment, companies need a peer opinion perspective to uncover their "why". We always start with a strategy session to align people, processes, and technologies toward the key outcomes.
  • Help your team to balance opportunity and risk: actuaries are one of the most trusted professions in the financial services industry. Beyond our quantitative abilities, we also bring acumen, innovation, ethics, and rigor to help you achieve balanced decision-making.
  • Customer obsession: we offer tailored solutions for your company that are innovative and cost-effective. Our experts bring leading insights and frameworks to modernise your team throughout the data maturity journey.

I am excited to join Michael Lip, Sammy Liu, and the rest of the Dataly Actuarial team in bringing more holistic offerings for both financial services and non-financial services companies.