Shuhei wanted an opportunity grow as a data scientist and software engineer. He found it at McKinsey.

August 3, 2021I joined Tokyo QuantumBlack, a McKinsey company, because I wanted to work in an international environment on projects that would make a difference and help me grow as a data scientist and software engineer. McKinsey has exceeded my expectations. I collaborate with international teams to help international clients. My work matters because I’m solving some of the toughest problems using data. I’m honing skills in software engineering providing top software to clients.

My most memorable project

I worked with a team of experts from Japan, Singapore and London on a project with a global B2B company. It was fascinating to learn which approaches worked best in different countries. The challenge we faced was to develop a software to optimize sales channels by leveraging data. This was easier said than done as artificial intelligence and machine learning projects are becoming more common, so projects can end up in the trial phase delivering insights but not actually creating value.

For this project, simply creating a machine learning model would not be enough. We needed actionable recommendations like how to leverage technologies to change how the client does business and “what if” analyses to target promotions that will maximize sales. To do this, we created a complex system in a relatively short time, combining several data science technologies such as supervised learning, causal inference, and mathematical optimization.

The benefits of a global team

This project made me proud to be part of such a talented global community. McKinsey invests heavily in creating and maintaining internal assets through QuantumBlack Labs, which meant we did not need to build a pipeline from scratch; rather, we started from our modular pipeline—a McKinsey asset—which provides a machine learning to optimization solution. Having worked at other consulting firms, I can say leveraging this asset is an only at McKinsey experience. Having the base pipeline freed our capacity for more creative tasks like combining multiple data sources for more accurate recommendations.

Support to succeed

We also received support from our global data science community. I was a little nervous about building an optimization model as a novice, but a senior consultant from London voluntarily taught me how to tackle the problem. This colleague coached me on how to build an optimization model, how to set up the objective function and constraints, including complex group constraints. When I struggled to implement some client requests, he spent time with me to problem-solve on the feasibility and implementation details. At the end, we implemented the optimizer successfully.

Measuring success

Our team delivered the analytics software within the project timeline and the software exceeded the client’s expectations. We accomplished this because we used machine learning models to understand the relationship between actions and expected sales, and used an optimization model to create actionable recommendations.

In the end, the software is easy for the client to maintain and scale. The software can easily be applied to another brand or product because of its flexible data platform, with building blocks Kedro to set up the pipeline in a logical structure and CI/CD to detect bugs when adding new features.

My journey with McKinsey has just started but I’m confident I’m on a great path.

More about Shuhei

Shuhei is a Tokyo-based senior data scientist at QuantumBlack. He holds a bachelor’s degree in technology management from the University of Tokyo. Prior to joining McKinsey, he was an actuary at Tokyo Marine & Nichido Fire Insurance Co., and a data scientist at Nomura Research Institute.

Outside of work, Shuhei enjoys karaoke, traveling (Pre COVID-19), and watching NBA.