Tokyo, Japan – Toshiba Tec Corporation and its subsidiary Gyainamics are working with McKinsey & Company, including its AI arm QuantumBlack, to advance how retailers use point-of-sale transaction data through artificial intelligence and GPU-accelerated computing. The initiative combines Toshiba Tec’s retail-data foundation, Gyainamics’ software engineering capabilities, McKinsey’s industry and applied-AI expertise, and NVIDIA GPU technology and software。
The work represents an important step in Toshiba Tec’s continuing transformation beyond traditional hardware. By applying accelerated computing and advanced recommendation models to retail data, the initiative is designed to support faster analysis, more relevant customer recommendations, and more timely, data-driven commercial decisions.
Unlocking the Value of Retail Data
For decades, Toshiba Tec’s point-of-sale systems have supported retail operations around the world, capturing large volumes of data from daily transactions. These data contain valuable signals about customer behavior and product demand. However, conventional CPU-based analytics have mainly been used for retrospective reporting, while real-time personalization and live profit optimization have remained difficult to achieve at scale.
In 2024, Toshiba Tec established Gyainamics within its business-building initiative to promote software development and create new digital business opportunities. Through Gyainamics, Toshiba Tec worked with McKinsey to strengthen the team and build core capabilities for a GPU-accelerated retail analytics platform.
The platform uses NVIDIA AI infrastructure and software to process retail data and support feature engineering, model training, and recommendation inference. The published case study describes the use of NVIDIA Merlin, including NVTabular and Transformers4Rec, together with RAPIDS cuDF and Dask.
Bringing Faster AI-Enabled Decisions to Retail
NVTabular moves data preparation and feature engineering from CPU-bound processes to NVIDIA GPU-accelerated infrastructure. This allows large volumes of transaction data to be normalized, encoded, and prepared for model training more quickly. The resulting data can then be used by transformer-based recommendation models.
Transformers4Rec applies transformer architecture to customer purchase sequences. It can learn from factors such as which products were purchased, in what order, how frequently, and in what basket or trip context. These patterns can be used to predict likely customer preferences and support more relevant product recommendations and promotional offers.
By shortening data-processing, training, and inference cycles, the platform can help retailers test promotions more frequently, strengthen recommendations across a broader portion of the product catalog, and adjust offers more quickly. The actual timing and business outcome depend on the deployment environment, data, operating model, and use case.
Building and Scaling AI in Production
A key focus of the work is not only building an AI model, but also operating and improving it continuously. Modern development environments, containerized workloads, and scalable machine-learning infrastructure help support repeatable development, production readiness, security, and ongoing optimization.
“Most enterprises face challenges not in building AI once, but in continuously operating and scaling it. Our partnership model with McKinsey solves that gap. That means clients access cutting-edge technologies – such as GPU accelerators and transformer-based recommendation systems – while ensuring that these tools remain production-ready, secure, and continuously optimized.”
Hiroyuki Koyama, President and CEO of Gyainamics
Reported Performance and Business Outcomes
According to the McKinsey case study, the GPU-enabled transformer approach produced significant improvements compared with earlier methods or historical baselines. The reported results include:
- A sevenfold improvement in personalization scores compared with previous recommendation methods.
- More than an 80% reduction in training time for large datasets.
- Expansion of product-catalog coverage from approximately 8% to 99.9%.
- Recommendation inference for 100,000 customers reduced from hours to under one minute.
- In controlled pilots, an average 5% lift in sales and profit per targeted segment compared with historical baselines.
- Up to a 7% improvement in long-term customer value compared with previous manual segmentation campaigns.
These reported improvements can allow retail teams to iterate more quickly, expand recommendation coverage, and obtain clearer visibility into promotional performance. They should be understood in the context of the specific datasets, methods, pilot conditions, and comparison baselines described in the underlying case study.
“Speeding up retail analytics using NVIDIA Ampere-generation GPUs marks a true inflection point. By combining Toshiba Tec’s deep retail-data foundation, NVIDIA AI infrastructure and software, and McKinsey’s applied-AI modeling, we’re demonstrating how advanced architectures such as transformers can reshape an entire industry.”
Takuya Kudo, McKinsey Partner
A Strategic Step Beyond Hardware
The initiative supports Toshiba Tec’s evolution from a traditional POS hardware provider toward a broader provider of intelligent retail solutions. It strengthens software engineering and machine-learning capabilities while helping the company explore software- and data-driven services alongside its established retail technology portfolio.
Gyainamics plays an important role in this transformation by incubating and scaling software initiatives. The collaboration with McKinsey helps build the organization, operating model, and technical capabilities required to develop and maintain AI-enabled applications. NVIDIA GPU technology and software provide the accelerated computing foundation used by the analytics platform.
Driving the Future of Intelligent Retail
As customer expectations and retail operating conditions continue to change, the ability to interpret transaction data and act on it quickly is becoming increasingly important. AI-enabled analytics can help retailers understand customer behavior, assess promotion performance, and make more informed operational and commercial decisions.
“The impact has been transformative. Retailers can deploy promotions with real-time ROI visibility, while consumer-goods companies gain transparency into how their trade spend is performing. This means no more spending trade budgets without understanding their true effectiveness.”
“Ultimately, we are creating a system that benefits retailers, manufacturers, and customers together.”
Hironobu Nishikori, President and CEO of Toshiba Tec
By turning checkout interactions into data that can support actionable insight, Toshiba Tec and Gyainamics are helping define a new direction for retail analytics. Together with McKinsey and supported by NVIDIA GPU-accelerated technology, the initiative demonstrates how retailers can move from retrospective analysis toward faster, more responsive decision-making.
Interested in connecting with us on retail solutions, co-creation, or partnership? Drop us a message or visit our website, we are ready to build the future with you.
Toshiba TEC Singapore Official Website
#ToshibaTEC #TeamToshiba