Smarter Retail: How AI Is Powering In-Store Optimization and Personalization

The retail sector is facing a radical transformation thanks to artificial intelligence. From in-store operational management to the personalization of the customer experience, AI enables data-driven decision making, anticipating demand and optimizing resources. At hiberus, we have developed advanced solutions as part of StoreFlow, which combines predictive analytics, occupancy modeling and real-world fraud detection to take in-store efficiency to the next level.

What is StoreFlow

StoreFlow was born as a comprehensive in-store management platform designed to address today’s retail challenges.

It is a flexible and scalable application that can be used in a wide variety of retail environments: from department stores to supermarkets to convenience stores. Thanks to its server-based architecture, StoreFlow provides the retailer with a centralized control platform, with real-time updates and constant availability of business status and sales information.

StoreFlow Key Metrics

The value of StoreFlow lies in its ability to translate operational data into useful, actionable metrics.

Key metrics measured by the platform include:

• Actual case usage: ratio of cases opened to cases actually used per hour.
• High occupancy: level of terminal saturation with respect to a predefined service threshold.
• Optimal opening: simulation of multiple staffing scenarios and selection of the most efficient combination.
• Lost sales: estimation of unattended sales opportunities due to lack of operational capacity.
• Anomaly detection: identification of irregularities such as unauthorized price changes, overwriting of alarms or atypical patterns in transactions.

Our contribution from hiberus

At hiberus, we have been responsible for the design, development and integration of the advanced analytics system that makes StoreFlow possible.

Using technologies such as Snowflake and PySpark, we built an infrastructure that allows us to analyze in real time the performance of the collection terminals. We have developed models that make it possible to anticipate occupancy peaks, plan new checkout openings and detect anomalous behavior using advanced predictive algorithms such as XGBoost, LightGBM or Isolation Forest.

New predictive models implemented

One of the pillars of StoreFlow is its ability to anticipate. These are some of the predictive models we have incorporated:

• Optimal opening scenarios (XGBoost + PySpark): dynamic prediction of the ideal number of cases opened, adapted by zone and customer profile.
• Returns forecasting (LGBM): segmentation of physical and online returns, taking into account seasonality, historical behavior and holiday calendar.
• Fraud detection (Isolation Forest): models trained with numerical and categorical data from various sources to identify suspicious transactions and set thresholds according to transaction type.

New in-store operational metrics

In addition to predictive models, we have incorporated more specific operational metrics for an even more detailed view:

• Identification of high occupancy cases and activation of additional terminals.
• Analysis of self-checkout (SCO) terminal performance.
• Estimation of missed sales opportunities.
• Advance planning for new openings or peak traffic.

Project Challenges

Every innovation presents challenges. In this case, the complexity of the retail environment posed a number of technical and organizational challenges.

One of the biggest challenges was adapting the forecasting models to such a dynamic ecosystem, with multiple cash fronts, customer types and workflows. It was also essential to achieve accurate operational forecasting, with very low margins of error, especially with regard to online and physical returns managed in the same flow.

To achieve this, we designed real-time processing pipelines that integrate data from multiple sources, without compromising accuracy or speed of analysis.

StoreFlow Benefits with our Integration

StoreFlow integration in physical stores has proven its impact with clear improvements in performance, experience and profitability.

• Significant reduction of queues and waiting times thanks to planning based on real data.
• Resource optimization with fewer checkouts open without affecting service quality.
• Increased operational security, with early detection of anomalies in sales and returns.
• Improved customer satisfaction, thanks to a more agile and frictionless shopping experience.

Project Results

Key performance indicators obtained after implementation reflect the benefits in a quantifiable way:

• -35% in high occupancy hours not resolved
• +20% efficiency in SCO (Self Checkout) planning
• < 0.1% false positive rate in detection of operational anomalies
• Returns forecast with more than 90% accuracy on a weekly horizon.

StoreFlow represents a new way of managing the point of sale, combining data, prediction and operational agility.

Thanks to this solution, we demonstrate that it is possible to make faster, more accurate and more profitable decisions even in complex physical environments. Every metric turns into action, every model into savings, and every alert into an opportunity for improvement. At hiberus, we continue to work to accompany our customers towards smarter, more agile and sustainable retail.

Do you want to harness the power of Generative AI to boost your retail business? We have a team of experts in Generative AI and Data who have developed GenIA Ecosystem, an ecosystem of proprietary conversational AI, content generation and data solutions adapted to the needs of the sector. We work together with the Retail and Distribution area to add value with technology, contact us!