In today’s rapidly changing world, it is becoming increasingly difficult to predict the future of our lives and businesses based solely on past experience and intuition. We often rely on media coverage or general perceptions when making decisions, but these can reflect biases or place too much emphasis on localized trends, making it difficult to fully understand what is really happening. What is needed today is decision-making grounded in objective facts and quantitative data. In other words, we need to look to the future based on data. Toshiba Digital Solutions aims to make data part of society’s infrastructure and create a sustainable future together with a wide range of stakeholders, including consumers, businesses, and local governments. Using proprietary AI technologies, we process and analyze purchasing data from more than three million consumers and combine it with transportation, weather, and other data sources to uncover hidden needs and emerging trends. Through this, we're helping create new value in diverse fields.

Market changes do not always align with what we see in the news or with public perceptions. Media coverage can be selective in its focus, and conversations on social media may amplify only certain voices. These sources reflect how society talks about issues, not necessarily how markets are actually behaving. By contrast, information that directly captures consumer behavior, such as purchasing data, functions as observed data that quantitatively reflects social trends. It provides a more accurate picture of changes in people’s values and demand. A perspective grounded in observed data allows us to better understand a rapidly changing society. To interpret real-world developments objectively, without being influenced by perceptions or assumptions, data backed by actual behavior is essential.

For example, as rice prices have risen sharply, recent commentary has suggested that consumers are “moving away from rice consumption.” However, consumer purchasing data collected by Toshiba Digital Solutions paints a very different picture (Fig. 1). Demand for rice declined temporarily in early 2025, but recovered following the release of government stockpiled rice and had returned to typical levels by June of the same year. Compared with the previous year, purchases of rice sold in bags weighing 10 kilograms or more decreased, while purchases of rice sold in bags weighing between 2 and 10 kilograms increased. The data suggests that consumers have shifted toward buying only the amount they need, even if that means paying a higher price per kilogram. Furthermore, purchasing data shows no significant shift from rice to bread or noodles. While it may appear that consumers are moving away from rice consumption, purchasing data suggests that what has changed is not the level of rice consumption itself, but the way consumers purchase rice.

In this way, data reveals both emerging signs of market change and the reality behind them. When making decisions in our lives and businesses, it is important to compare multiple sources of objective data and interpret them accurately. Rather than accepting the information we encounter every day at face value, we need a more complete picture of how society is actually moving, grounded in data.

The value that any single dataset can provide is inherently limited. However, that value can be greatly enhanced by enriching data with metadata and integrating different types of information, such as purchasing data, mobility data, population statistics, and weather data, and by understanding the relationships between them. In other words, data that may appear unrelated at first glance can be connected to form a kind of map that helps us understand how society is moving.

Consider, for example, mobility data showing that a person traveled from Point A to Point B. On its own, that data cannot explain why the trip was made. However, by linking it with purchasing data, such as ticket purchases for a live event, shopping activity, or restaurant visits, it becomes possible to understand the purpose of the trip and whether it was a one-off event or part of an ongoing pattern. Insights into whether movement patterns are temporary or represent sustained changes in foot traffic can help inform rail service planning and improvements to facility operations. By gaining an integrated view of human behavior and changes in communities, organizations can uncover opportunities for new business development and the reevaluation of existing business models. One of the greatest benefits of data integration is the ability to reveal the underlying structures and context that cannot be seen when data is viewed in isolation.

However, in Japan today, much of the available data remains confined within individual companies and organizations, limiting the value that can be derived from it. If mechanisms for enabling data sharing across industries and domains become more widely adopted, the value of data will increase, allowing products and services to become more sophisticated and intelligent. Our goal is to position data as the foundation of the products and services that support society. In other words, we seek to make data part of society's infrastructure.

This approach aims to extend the framework for data utilization across society, enabling everyone to use data as a shared foundation. Starting with purchasing data generated by consumers and combining it with a wide range of other data sources, we aim to create a platform that supports decision-making throughout society. In an age characterized by an abundance of information and choices, we believe this platform will contribute to more reliable and informed decision-making.

Figure 2 provides an overview of the platform we envision, showing how data is collected, integrated, analyzed, shared, and utilized.

The volume and quality of the purchasing data that serves as the foundation of this platform, together with the technologies used to process, classify, analyze, and visualize that data, are key elements of the data-sharing ecosystem that follows. The Toshiba Group offers two smartphone applications that support the management of consumer purchasing behavior and purchasing information. One is Smart Receipt, which allows users to receive receipts electronically instead of on paper. The other is Receipt Scan, which enables users to digitize paper receipts using a smartphone camera. By using Smart Receipt and Receipt Scan together, users can easily digitize and centrally manage receipts from stores that are not part of the Smart Receipt network.

These two services serve as entry points for recording everyday purchases and are distinguished by their ability to digitally collect and centrally manage a wide variety of purchasing data. Together, they have more than three million users, and Smart Receipt is used by 17,657 stores operated by 573 companies, including supermarkets, convenience stores, drugstores, home improvement stores, co-ops, and restaurants.* Because of this scale, the data can provide a highly accurate picture of market trends.

*User figures are current as of November 2025. Store participation figures are current as of June 2025.

Data collected through Smart Receipt and Receipt Scan is accumulated on our big data platform as purchasing data. One of the defining characteristics of the platform’s purchasing data is that it combines both depth of consumer insight and large-scale data volume. Generally speaking, purchasing datasets tend to excel either in depth or in volume, but rarely both. For example, ID-POS data (a combination of point card ID data and Point of Sale (POS) data) collected by retailers may offer large volumes of information, but because it only captures purchases made at a specific retailer, it provides a limited view of consumers. Household budget data, on the other hand, offers rich insights into individual purchasing behavior. However, because it is collected from a relatively small number of people, it is difficult to use for large-scale quantitative analysis. Toshiba’s purchasing data achieves a balance between both depth and volume.
Because Smart Receipt and Receipt Scan data are managed under a single customer ID, it is possible to gain a holistic view of purchasing behavior across multiple retailers for more than three million consumers. In particular, the ability to capture consumer behavior across multiple stores is a unique strength that offers insights unavailable from the data of any single business.

Analysis on this platform is conducted using anonymized data that cannot be used to identify individual consumers or stores, in accordance with user permissions and agreements with participating retailers. For example, the data might indicate that “a woman in her thirties living in Prefecture A purchased three 500-milliliter bottles of water at a supermarket in Prefecture B at 10:10 a.m. on January 1, 2026, at 148 yen per bottle.” Data such as this is processed and analyzed using AI technologies specialized for receipt data to identify hidden needs and emerging trends. The resulting insights are then provided to a wide range of customers as a basis for decision-making.

Purchasing data generated through Smart Receipt and Receipt Scan is already being used in a wide variety of fields. Consumer goods manufacturers use it for demand forecasting, financial institutions use it to support stock market forecasting, and local governments use it to analyze regional trends. One of the platform’s greatest strengths is its ability to create value across a wide range of industries through a single data platform.

To efficiently process and analyze receipt data and transform it into high-value information, Toshiba has developed its own AI technology specialized for receipt data, known as Receipt Informatics. Receipt Informatics is a suite of AI technologies that includes clustering AI, which uncovers hidden purchasing trends by grouping products and identifying meaningful analytical dimensions; graph analysis AI; and large language model technologies that accurately infer appropriate product categories from item names printed on receipts.

Products sold in stores can generally be divided into those managed using standardized JAN codes and so-called in-store products, which are managed using retailer-specific codes instead. In supermarkets, for example, fresh meat, fish, vegetables, and prepared foods fall into the category of in-store products and account for roughly 40% of purchasing data. Because these products are difficult to classify in detail, a significant portion of the data associated with them has traditionally gone underutilized. To address this challenge, we developed technologies that identify product categories from the names printed on receipts, enabling this information to be used as actionable data.
As a result, we can now comprehensively classify products, including those without JAN codes, making it possible to capture a complete picture of purchasing behavior without gaps.

In addition to these AI technologies, we have built a big data platform using AI technologies cultivated through manufacturing applications, state-of-the-art AI technologies, and advanced big data processing technologies designed to improve efficiency and reduce energy consumption. By leveraging multiple AI technologies on this platform, we provide analyses that help customers refine and enhance their sales strategies.

We also place great importance on data reliability. In the case of Smart Receipt, once data has been incorporated into the big data platform, neither users nor participating retailers can arbitrarily alter it. As a result, it can be used as reliable proof of purchase, including for documentation submitted under Japan's Self-Medication Tax Deduction program.

Naturally, data governance is also managed with great care. As noted earlier, all data is used in a manner that prevents the identification of individuals or specific stores. We have also established an external advisory panel on privacy protection that regularly reviews data utilization from an independent perspective. Ensuring both transparency and reliability is an essential element of our data platform and allows users to use it with confidence.

Analysis powered by our purchasing data and AI technologies has been recognized by a growing number of companies and organizations for its ability to reveal real consumer purchasing behavior, and its applications continue to expand year by year.

For example, ZENB JAPAN, Inc. faced the challenge of measuring the size of the gluten-free market because there was no established product category for gluten-free products. By applying our clustering AI technology, the company identified purchasing preferences closely associated with consumers of gluten-free products, including low-carb, high-protein, and high-fiber preferences (Fig. 3). This made it possible to uncover connections between consumer purchasing characteristics and untapped market potential. ZENB JAPAN has praised the technology for making it possible to visualize relationships that had previously been fragmented across multiple product categories and difficult to analyze holistically. The insights gained have enabled the company to pursue market development strategies, including cross-category marketing.

Analysis of this kind, which reveals previously unseen market relationships, also contributes to improved product development and brand strategy.

We are also collaborating with NTT DOCOMO, INC. in the field of marketing solutions.
By linking Smart Receipt with NTT DOCOMO’s d ACCOUNT, we are helping deepen understanding of customer behavior based on purchasing data and supporting more sophisticated marketing initiatives. This is one example of how purchasing data can fill the "last-mile" gap in understanding customer behavior and improve the precision of digital marketing.

In addition, our services are being used by local governments across Japan in initiatives related to regional revitalization, public health, and the environment. For example, purchasing data related to food is being used in research that identifies dietary trends within communities and analyzes dietary patterns that contribute to longer healthy life expectancy. We believe that supporting the development of evidence-based measures for a sustainable future represents an important contribution, both for businesses and for the public sector. The use of data generated through this platform to visualize regional issues and support local government policymaking is also continuing to expand.

We see AI agents as a core technology in our efforts to integrate multiple types of data, starting with purchasing data, and to gain a deeper understanding of people and communities. While conventional AI has excelled at processing individual datasets, it has had limitations when it comes to understanding the broader context in which multiple factors interact across purchasing behavior, mobility data, demographic attributes, and regional characteristics. In other words, it has not been easy to infer the intentions behind people's behavior or the meaning of their actions from data alone.

Our goal is to realize two types of AI agents: human-proxy AI agents that can reflect human behavior and intent, and human-augmentation AI agents capable of highly accurate, bias-free analysis. Through these two types of AI agents, we aim to make more effective use of both purchasing data derived from individual behavior and objective infrastructure-related data such as real estate, population, weather, and railway passenger-load information, while preserving the context in which that data exists.
For example, by combining purchasing data, mobility data, and user attribute data, we can generate information that is useful to individuals. Similarly, by combining purchasing data with real estate data, we can assess regional potential and provide reliable insights backed by data. Furthermore, the platform is designed to analyze infrastructure-related datasets, including communications, transportation, and energy data, from multiple perspectives. By doing so, it can provide insights that support decision-making by businesses seeking to improve convenience for consumers through initiatives such as tourism recommendations, tenant attraction strategies, and the smoothing of peak traffic flows (Fig. 4).

In Japan, industry, academia, and government are also working together to build and expand data spaces that enable organizations to share data across organizational boundaries while retaining control of their own data. These data spaces serve as social frameworks for circulating data, while AI agents function as the intelligence layer that understands context and creates value within those frameworks.

Based on our experience supporting Japanese manufacturing and digital transformation (DX), we believe we can contribute both by creating mechanisms that balance data protection and data collaboration and by deploying AI agents that operate within those frameworks.

Starting with purchasing data, AI agents can develop a deep understanding of context across multiple datasets and, through data-space frameworks, deliver that value to society as a whole. Our mission is to evolve data into part of society's infrastructure and create new value together with companies, local governments, and consumers.