Financial engineering is considered an elite profession. If you are asked to envision a financial engineer, you imagine fancy White collar folks in Wall Street offices leveraging advanced algorithms for trading. And that is somewhat close to reality. Everyone will label them as “numbers people.” After all, what drives trading is numbers and algorithms. Right?

Now if I ask you to imagine a warehouse operations manager, you may find that imagination less exciting. In most people’s minds, it is about managing operations in the warehouse. Moving forklifts, pallets on racks, picking, trucks going in and out-pretty much that. Most individuals who do not work in the supply chain will not associate warehouse management with fancy algorithms.

They are not entirely wrong in their assumptions.

We do not run warehouses leveraging as much science as we leverage in financial trading.

A decade or two ago, you could argue that the benefit of leveraging advanced algorithms for managing day-to-day warehouse operations was not worth the investment in developing advanced algorithms.

That is no longer the case.

There are algorithms from financial engineering that you can lift and apply directly. Others, you can tweak a little bit. I am sharing one example in this article, which I believe is the most straightforward application. As I advance my learning in the area of financial data science, I see opportunities that can be game-changing.

And maybe, one day, a senior at a fancy college will dream of managing a warehouse driven by science, algorithms, and robotics!

Example of applications in finance

Recurrent neural networks (RNNs) are networks designed to learn about sequential data, like text data or time series data. We will use one example of application of RNNs in finance, specifically prediction of prices and price movement.

Price and return series data: As mentioned in the introductory section, RNNs are good fit for time series data. Hence financial time series data is a good candidate for RNNs. An example is currency trading. Intraday currency quotes (Ex: INR/USD). RNNs can leverage this type of time series data to make predictions.

Predictions need not always be estimates. With binary labels, you can leverage the model as a classification model, predicting the direction of the movement of the price or return.

Extrapolating applications to the supply chain world

Let us extrapolate the same RNN application to warehouse operations management, specifically Sales and Operations Execution (S&OE).

Day-to-day management of warehouse operations is not as straightforward as you may imagine. Many planning elements are involved in weekly and daily warehouse operations planning. That is what Sales and Operations Execution (S&OE) is about.

Unfortunately, warehouse managers are so swamped that they mostly use their expert knowledge and brute force to plow through daily and weekly challenges. There is room for science. The warehouse exists to hold and move products to cater to demand. As demand fluxes, so should warehouse operations. With a lack of sound science, this need to flux with demand often introduces chaos.

If you imagine the type of data that you capture at weekly and daily levels, you have essentially a time series. Let us say you want a view of what your next few weeks may look like, at a day-to-day level, at the SKU level. The SKU-level picture often becomes essential. For example, bottled water, an SKU with a low-profit margin in consumer goods, hogs a significant amount of warehouse space. You do not want to carry a certain amount throughout the year since the space it will consume may be at the cost of SKUs with higher profit margins.

Leveraging RNN, trained on at least five years of SKU-level data, you can gain a level of prediction that conventional forecasting methodologies will not provide. Since RNNs perform better with large data volumes, you can add layers like warehouse slotting constraints as well in your model. As mentioned previously, this is just one example and the simplest one.

Algorithms that can transform and streamline warehouse operations already exist. We are not looking hard enough.

References:

• AI in Finance- Yves Hilpisch, O’Reilly
• Principles of Financial Engineering- Kosowski, Academic Press
• Handbook in Monte Carlo Simulation: Applications in Financial Engineering, Risk Management, and Economics- Brandimarte, Wiley
• Practical Financial Optimization: Decision Making for Financial Engineers-Zenios, Wiley