After three years of struggle, what is the ideal engineering organization that XICA's architects are aiming for?

There were about three really tough times, but now I'm in the best place by far.

That's what Shintaro Matsushima, an architect at XICA's Development Headquarters, told us. Matsushima studied mathematics and computer science at university, and started a company while he was a student. His work in development didn't lead to business, so he turned to freelancing to improve his adaptability and flexibility as a person. It was during this time that he encountered XICA and joined as a contract worker. Five months later, he became a full-time employee as an engineering manager, and is currently in charge of re-architecture and the mathematical aspects of products. We asked Matsushima, who has overcome many difficulties, about his thoughts on the ideal engineering organization.

Three years of thinking, "If only I just tried a little harder..."

-- It's been three years since you joined the company in July 2017. I think the company and organization have changed a lot.

Of the engineers who joined the company, I am the last remaining one.

Three years ago, the difficulty of the product made it difficult for engineers to maintain their motivation. It required highly advanced statistical knowledge and understanding of the customer's business, and the implementation was not in a good state.

── But how have you managed to continue this far?

Amidst the chaos, I felt there were many elements of success.

First,There are many passionate and sincere peopleRegardless of department, everyone had the desire to succeed. There were many areas where we were not doing well, but we were not blamed for that.A team that helps each other by asking each other, "Is there anything we can do to support each other?"

On the product side, the code was indeed a mess and difficult to read.
But the foundation we've built up in the year since the release isn't so bad that we should throw it away. We can do better technically. We're setting high goals, but I don't think they're just pie in the sky.

"We're pretty close to success. We just need to work a little harder and we'll be successful."I've been working with that mindset for the past three years.
It was quite difficult when I tried it though (laughs).

If the product disappeared now, XICA would have nothing to sell. We felt a sense of responsibility because we are the engineers who make it.I wanted to somehow connect the product to the future as an engineer's value.Hm.

── Just from this story, you can get a glimpse of the hardships he endured over the past three years. Mr. Matsushima joined the company as a contractor, and became a full-time employee as an engineering manager five months later.

I had been getting work from acquaintances since my university days, and as an extension of that, I started a business with three friends.
I tried my best to create a lot of different things, but they didn't turn into business and I got burned out, so before joining XICA I worked as a freelancer on various development projects as a form of training.

I had gained a fair amount of experience as an engineer, so I had a certain degree of confidence in the technical aspects.
But,What value do I have when I can't compete on technical ground?I joined XICA because I wanted to test my adaptability and flexibility as a person and see if I could do something even though I can't write code.

If you keep doing the same thing, your learning curve slows down. But if you try something you've never done before, the initial improvement is incredible. I wanted to gain that experience.

When I was approached about being an engineering manager,Expecting discontinuous growth for yourselfAccepted.

The troubles of being a manager

── It seems like you had a lot of hardships during your time as an engineering manager.

Yes. At the very least, I couldn't drive the company organization and business. When I was a manager, we had to pivot (※1) a lot at the stage of whether the current product would sell or not. Every time we pivoted, the engineers would cry out in frustration, bugs would appear, problems would occur, and we would have to apologize. It was a repeated cycle. I wasn't able to communicate with the team members about the fundamental story of "what kind of product it is" and "what we're doing with the statistics."

But I was still deeply moved when the members told me that they enjoyed working here.

── "I want to help CS" "I don't want to cause any trouble to customers"

Even though they caused bugs and made the code dirty, everyone was facing the project with a sense of responsibility. Even if they were not very good at technical things, they contributed to the product with their efforts and passion. That is why I always regret that I was not able to present their careers as a manager and encourage their further growth.

Only when the architecture is neatly put together will you look back on the accomplishments of your colleagues and the connections you have made, and think, "Ah, that was a great place."Passing on the baton to them is the pride of being the last remaining engineerI think it is.

*1: In product development, the direction of the business or product changes to meet market needs, and the product is improved accordingly.

Re-architecture and mathematics

── Currently, you are in charge of rearchitecture and the mathematical aspects of the product. What exactly does that involve?

As an architect, I am responsible for system design of the entire product.

We are currently proceeding with the overall design, including how to make it last for three years and how well it can withstand future expansion, with review by members, management, and advisory board members Taku Namura and Shinji Tanaka.

As a mathematical engineer, I am in charge of implementing specialized statistical knowledge into the product. Murata Kenta (mrkn) joined the project and we put together the "statistically desirable functions" using mathematical logic, organized and summarized the feasibility and implementation patterns, and then implemented them. The mathematical part starts from very academic content. I am in charge of the conversion part, which incorporates academic elements into the product.

Once the re-architecture is accomplished

── What will be possible once the re-architecture is achieved?

The current challenge for Magellan is that the application and mathematics are tightly coupled. The hurdle is high because only people who understand both mathematics and development can understand the product structure, and since there are few people who have experience with it, it is easy for it to become personalized.

There are three expected benefits to separating applications and mathematics through re-architecture.

one is,Accelerating development speed.
This will allow us to update mathematical functions and conduct experimental challenges more quickly, making R&D (※2) easier. Since it will be accessible even to those without knowledge of both mathematics and development,Eliminating personalization will be important.

The other isCreating synergy between development and the business sideIt is that.
Currently, it is a one-way situation where we create the features that the business side wants, but I think that in the future engineers will be able to make proposals to the business side.

Finally,Customer Success (CS) can now focus on what they do best, and better advertising models can reach clients.The idea is to become like this.

Currently, the flow is that what is created by development is optimized and used by CS, but that can be resolved by implementation. I would like to improve non-essential aspects with this re-architecture so that fitting and comparing models will be easier.

*2: Research & Development: Activities such as research into a company's business field, development of new technologies, and technological research and development necessary to increase the company's competitiveness.

thoughts on the product

── You must have experienced the difficulty of dealing with specialized statistical techniques many times in the past, but what do you think of the Magellan product?

When I joined the company three years ago, machine learning and deep learning were the polar opposites of statistics. With AI attracting attention as a technological trend, I think it was the right decision to consistently choose statistics as my tool.

AI analysis involves tens of thousands of explanatory variables. Because it is an automatic calculation using an algorithm, the analysis part is a black box. Recently, machine learning and deep learning have been trying to improve their explanatory power, but humans still cannot fully understand the reasons why something happens.

When placing ads, marketers create a story of "how to lead customers to conversion." We are dealing with clients' huge advertising budgets, amounting to tens or hundreds of millions a year. Failure is not an option.

At XICA, our CS experts have used their statistical know-how to model analytical reports and budget simulations that clients use for decision-making.Statistical analysis, which allows us to visualize and explain all analytical processes, is the right approach to the value we want to provide.I think that's how it was.

── The value of this product lies in the fact that it makes everything visible and allows humans to understand the entire process.

An organization that began experiencing rapid growth

── What are your thoughts about the changes in the company and organization over the past three years?

I think my teammates from three years ago would not be able to believe what's happening now. They'd be like, "Is Magellan really this cool?" (laughs)Today's organization is a miracleWhat is it?

In January 2020, Korezawa-san was appointed CTO and everything changed. It has changed so much in just 1 months that it feels like a different company. On the other hand, I am not sad that my colleagues who I used to work with are gone.I feel so changed that I can't even remember the past.This is exactly what"Discontinuous growth"it might be.

We had many talented people join our company, many of whom we would have had to hire over the course of three years.The organization is also undergoing a discontinuous build-up.I'm very grateful to Koresawa-san.

But I think we're still at the starting point, so I want to keep going at it. I was hoping for a discontinuous rebirth for myself and the organization, so I'm very happy.
So,The current XICA is by far the best.

If I ever see a reborn, standardized Magellan from now on, I might die (laughs).

── If Matsushima says that much, surely the engineers who will join his team will be incredibly happy?

Isn't it good to be so happy? It would be great if the company could grow stronger and stronger.

Magellan as it is today is the product of the combined efforts and passion of our young engineers. I want to carry that on into the future, and it would be fantastic if we could make it even more beautiful.

I think standardization is about transforming the way things are presented or combined in a way that brings out value. It would be great if the re-architecture I'm doing could create an environment where new people can thrive.

*The affiliation and title of the interviewee are those at the time of the interview.