Sometimes it helps to look at our OR industry from outside. This is the story of a tax specialist and the challenges he encountered while trying to use mathematical optimization in his professional area.

Over the course of my career in financial services, I’ve witnessed firsthand both the promise and the pitfalls of mathematical optimisation (MO). Here are some key lessons and insights from my journey.

Part 1: The Spark

My introduction to MO came during my financial services training, where my studies included operations research and linear programming. Years later, whilst I was working on a task that involved complex decisions I wondered: Could optimisation techniques help? I experimented with Excel’s Solver and soon built a model. The results surpassed what I had produced manually, and the process was surprisingly easier to implement than I expected: it didn’t require advanced expertise or a steep learning curve.

Key Takeaways:

• Awareness is Key: I only used Excel’s Solver because I knew about linear programming. Awareness of MO is essential for individuals or firms to recognise when a problem can be addressed by MO.

• Accessible Tools: Excel’s Solver is available and free in Office365. It’s quick to learn, and deep mathematical knowledge isn’t required.

• Low Barrier to Entry: With little time and no extra expense, anyone can experiment with MO and quickly judge if deeper investment is warranted.

• Limitations Exist: Entry-level tools can have limitations, such as scalability, performance, and the risk of finding only locally optimal solutions. As needs grow, more specialised tools and techniques may be required.

• Learning Curve: Starting with Excel is a great entry point, but users will encounter limitations that signal when it’s time to explore more advanced solutions.

Part 2: Lessons from Large-Scale Implementation

In a subsequent role within a large financial services firm, the use of MO was considered. A business case was developed and approved, but the journey was faced with unforeseen challenges.

So, what went wrong:

• Consultancy mismatch: A consultancy recommended a tool ill-suited for our needs, leading to performance issues.

• Inefficient model: A consultant’s formulation for a critical part of the model used unnecessary binary variables, that consumed excessive memory and severely affected performance.

Attempts to improve:

·       We consulted additional experts and explored strategies. While some improvements were achieved, various constraints limited our ability to make further improvements.

Outcomes and Lessons Learned:

• Awareness gap: Even among experienced professionals, knowledge of MO and its potential was limited.

• Due diligence: It’s important to thoroughly evaluate external partners and solutions, and to seek multiple perspectives for high-value or high-risk projects.

• Testing: Robust and varied test data is essential to identify potential issues early.

• Collaboration: Broader consultation and open-mindedness can lead to better outcomes

Part 3: Building a Better Solution – and Uncovering New Challenges

Later, having left the financial services firm, I developed a new application using more advanced optimisation tools and benefited from collaborating with an expert who provided an extremely efficient linear formulation for a complex task.

So, what happened:

• The resulting application was powerful and efficient, impressing many people.

• MO’s disruptive potential: The tool revealed inefficiencies in existing systems, sparking both excitement and concern.

• Beyond "business as usual": It enabled workflows previously deemed impossible.

Broader observations:

• The awareness gap persists: even technically skilled professionals often don’t know MO exists - or fear it’s too complex.

• The perception that MO is only for mathematicians or data scientists is a barrier, but in reality, it is accessible to a much wider audience.

Final Thoughts and Suggestions for Optimization4All

• Guidance: A clear, easy-to-follow flowchart would help users avoid common pitfalls and challenges.

• Onboarding:  Pre-prepared materials and highlight free, accessible tools for beginners.

• Demystification: The term “mathematical” in MO can be intimidating. It could be helpful to convey MO is not just for mathematicians or data scientists. I’m neither, yet I’ve become proficient.

• AI Assistance: Mention that generative AI can now help build models, making it even easier for newcomers to get started.

How Optimization4All Can Bridge the Gap

My story underscores a widespread lack of awareness, practical guidance, and accessible resources for MO-especially outside traditional optimisation circles. Optimization4All can address these gaps by:

• Raising Awareness: Share real-world case studies and success stories.

• Educational Resources: Offer step-by-step guides, tool comparisons, and beginner-friendly materials.

• Expert Directory: Provide a vetted list of consultancies and experts for second opinions and support.

• Community: Foster a collaborative space for sharing experiences, lessons, and best practices.

• Demystification: Make it clear that MO is accessible to all professionals, not just mathematicians or data scientists.(MO).