Collaborative Ecosystems with Data and AI at Rakuten Institute of Technology

, Masaya Mori, AI

New Approach to Innovation – A Collaborative Ecosystem

    Data is one of the most important resources for the economy in the present age. And what’s more, AI is now bringing new value to society at a significant pace.

    Before getting to the subject of data utilization in Rakuten, I’d like to look back on the passage of time. It’s a coincidence that we can describe the development of civilization to date using four “I”s.

    First, the era of “Industrialization” was brought about by the industrial revolution in the 19th century. The era of “Information” powered by the popularization of computers followed that. The third era is the important one, “Individualization”. In the past couple of decades, as we have moved through the advent of the internet and social networking services, technology has allowed people to express their diversity more clearly. Different people basically always have different interests, so internet services have been focusing on developing personalized functions, and we believe we can utilize AI technology to provide different services more suitable for each individual. But, technology is evolving at a significant pace, and new disruptive solutions are emerging one after another. Now, we’re facing the new challenge that we need to create new business dynamism in order to make the most of new solutions including state-of-the-art AI. That is the era of “Innovation”.

    Rakuten was founded in 1997, and we have started a variety of services which span “Rakuten Ichiba”, one of the world’s biggest online shopping malls, an online travel agency, digital content, communications, fintech such as bank, securities, insurance and e-money, and professional sports, bringing the joy of discovery to more than 1 billion members across the world. While we have now over 70 businesses and we’re growing and growing, our way to launch new services is also changing.

    Looking back over the past 10 years, new services tended to be developed mainly by web engineers. However, currently this is shifting from web engineering-centric to startup-centric services. Groundbreaking services now are created based on the idea of startup companies. Let me take an example. The founders of Google renewed the strategy and the company formation to establish the holding company, Alphabet, and proceed with venture businesses like autonomous cars under it. They are now bringing about new businesses by driving individual startups. You see this kind of trend not only in Google, but in many companies globally.

    On a related note, the Rakuten group changed its mission statement from “To become the No 1. internet service company in the world” to “To be a Global Innovation Company”, expressing our new resolve to actively launch new businesses. We also think that interaction with existing businesses plays a significant role in starting an innovative business.

    What we may want to pay attentions to is, while startups can experiment with totally new projects, they don’t have so much data to work with. That’s why it is important that they partner with data platformers who have a well-organized management system of data utilization. On the other hand, some old enterprises have a huge quantity of data, but they don’t have as much capacity for generating ideas and innovations as startups do. Naturally, it is a very significant strategy to connect both of these.

Rakuten open data for academic research

    Rakuten Institute of Technology provides a variety of data held by the Rakuten group to universities and public research institutions for research purposes (cf. https://rit.rakuten.co.jp/data_release/ ). Our open data mission consists of three parts: contribution to the development of applied technology in academic fields; acceleration of the evolution cycles in technology by strengthening links between academia and enterprises; and promotion of unique research by holding symposiums and supporting application development using large-scale data. To date, more than 250 universities, laboratories and research institutes have used the data for their research, and Rakuten Institute of Technology reflects their results in our service development. It is vital for us to collaborate with academic organizations in order to generate a variety of businesses.

    Rakuten has been growing in the e-commerce field since its foundation. In the e-commerce business – where you don’t meet customers in person – data is the only measure you have to understand them. I think that the data which Rakuten has accumulated over the past 20 years has a strong potential for utilization in many areas. While we never disclose our confidential data to external parties, we understand the significance of open data and are careful in providing data for research purposes.


Rakuten Data Release Website: https://rit.rakuten.co.jp/data_release/

Data utilization is a way to survive in a competitive world.

    Rakuten Institute of Technology was founded as a research organization in 2005 to predict the trends of the internet and create new technologies. It has more than 120 members in 5 locations: Tokyo, NY, Boston, Paris and Singapore. The researchers –most of whom have a PhD in Computer Science – conduct a broad range of research on topics such as AI, IoT, drones and AR/VR based on their own vision to reflect their results in the business of Rakuten.

    Back before we launched our research institute, we had intensive discussions with many professors in universities as well as researchers. We found out that they had strong concerns that their technologies were not going to be used in live business situations even if it had groundbreaking functions. That is, they were afraid that their research would end up as nothing more than words on paper because they didn’t have much chance to look at actual data or to discuss real issues with business people. To respond to these challenges, we established our own unique style where we focus deeply on data, create opportunities for our researchers to have continuous dialogue with business people working on the frontline, have discussions on issues combining viewpoints from business and academia, share goals, and proceed with research together.

    Rakuten researchers interact and collaborate globally with many of the Rakuten group’s 70 plus businesses. The business side knows the state-of-the-art technology and the latest academic methods. They come to Rakuten Institute of Technology after identifying business issues and discuss how to overcome them with researchers to lead to practical solutions for business. We are working every day to deliver meaningful results.

    Let me share some examples of implementing AI technology in the Rakuten group. We developed a trend analysis technology to predict the sales of each product by making the most of marketing data of more than 250 million items. Some products were thought to only sell in one specific season of the year. However, this system found another potential season when product demand spikes unexpectedly. For other products, the system found correlations between sales and seasonal events which don’t appear relative to each other, by analyzing the time series data automatically. It was an unexpected find that some products sell only in accordance with Father’s Day or Mother’s Day. You could obtain such findings through manual analysis if there were only several products. But, humans cannot cover anything like all of the 250 million items. Our AI can analyze such high-volume data to learn about potential needs. This system was made simple and now marketers and EC consultants all over the world in Rakuten are using it for strategic planning or for providing consulting services to merchants in our EC market place.

    Furthermore, we leveraged image recognition technology with deep learning on Rakuma, our C2C marketplace app. The user takes a photo of an object that they want to sell and activates the function “Moshikore” (meaning “take a guess of this”) which is a category classification. It will recognize what’s in the photo and automatically recommend the category that the object should be assigned to according to the confident score. This function enables the user to save time registering product information when they want to sell something.

    Recently, we’re paying more and more attention to fintech businesses and adapting AI prediction technologies to finance markets. We even tried to predict the economic market in Japan and the results from our model were less than 0.1% different from indicators announced by the government. We already leveraged it to investment activities using our bigdata and supporting the business management of merchants.

    The reason why Rakuten links data science to business and promotes data utilization is because we firmly believe it’s the only way that we will survive in the global competitive environment. So, we’re going to collaborate with more and more startups, companies and academia by sharing data and utilizing AI.