Tag: Artificial Intelligence

Amidst the pandemic-fueled surge in online shopping, porch piracy emerged as a prevalent concern, with over one in 10 adults falling victim to package theft within the previous year, according to a 2021 Consumer Reports survey. This modern-day menace, epitomized by the term “porch pirate,” underscores the vulnerability of unattended packages to opportunistic thieves. UPS Capital recognizes the challenges faced by its customers in securing their package delivery ecosystem and is harnessing digital capabilities and data access to redefine traditional approaches, ensuring improved customer experiences and combating shipping loss.

About UPS Capital

UPS Capital, a subsidiary of UPS, specializes in providing financial and insurance solutions tailored to businesses engaged in shipping and logistics. Established in 1999, it offers risk management services including cargo insurance and trade credit insurance, along with trade finance solutions such as supply chain finance and export financing. UPS Capital provides customs brokerage services to navigate import/export processes, supply chain optimization tools like supply chain analytics and inventory management, and technology solutions like the UPS Capital Merchant Services platform and UPS Capital Cargo Finance platform. These offerings collectively support businesses in mitigating risks, optimizing operations, and facilitating smoother transactions within the global trade and logistics landscape.

Challenges

The surge in online shopping has led to an unprecedented rise in package deliveries and, correspondingly, package theft. This upsurge has dramatically outpaced traditional security measures, exposing significant operational vulnerabilities within UPS Capital. The sheer volume of data generated from the increasing package deliveries overwhelmed existing data management systems, underscoring a critical need for more advanced data handling capabilities. The absence of real-time data processing capabilities hindered UPS Capital’s risk management and rapid response efforts. This deficiency affected not only operational efficiency but also eroded consumer trust and impacted the financial performance of the company. These multifaceted challenges highlighted the urgent need for a sophisticated solution capable of addressing the complexities of modern package delivery and logistics.

Solution

In response, UPS Capital integrated Striim’s real-time data streaming technology with Google BigQuery’s analytics capabilities to enhance delivery security. Striim’s platform enabled the immediate ingestion and integration of data from various sources, facilitating real-time risk assessments and proactive decision-making. This seamless data flow into Google BigQuery allowed for advanced analytics, leveraging AI and machine learning to predict potential delivery risks and optimize logistics strategies effectively. Additionally, the innovative DeliveryDefense™ Address Confidence system utilized this integrated data to assign confidence scores to each delivery location based on real-time and historical data, enhancing predictive accuracy. This system empowered businesses to proactively manage delivery

risks by rerouting packages or adjusting delivery protocols based on the calculated confidence scores, thereby streamlining operations and enhancing security.

The UPS DeliveryDefense program utilizes a sophisticated technical setup, starting with the direct upload of varied datasets into BigQuery. This platform acts as the primary structured data repository in Google Cloud. Concurrently, SQL Server data is thoroughly cleaned in Link Data, which also extracts images and email attachments from different systems, ensuring data integrity and availability. These enriched datasets are merged in BigQuery for seamless Google Cloud integration. Vertex AI then becomes pivotal, running advanced machine learning models like route anomaly detection and fraud detection for shipping transactions. Using Vertex AI’s extensive tools, these models are trained, refined, and implemented to discover insights, predict trends, and extract valuable information. Firestore, a flexible database suitable for various development environments, stores insights, confidence scores, and analytical details, all accessible via the Looker API.

Results

• Improved Customer Experience: The integration of Striim not only secures deliveries but also optimizes routing and delivery strategies, resulting in heightened reliability. This reliability, in turn, boosts customer trust and satisfaction, as customers receive their packages safely and on time.
• Cost-Savings: UPS achieved significant cost reductions by implementing advanced strategies to minimize losses from theft and optimize delivery routes, employing proactive risk management alongside sophisticated analytics and route optimization algorithms.
• Advanced AI and ML Implementations: Utilizing Striim in conjunction with Google Cloud technologies like BigQuery and Vertex AI, UPS can deploy complex machine learning models. These models are crucial for detecting routing anomalies and preventing shipping fraud, thereby enhancing the security and efficiency of the delivery network.
• Improved Data Processing and Analytical Accuracy: Striim’s implementation of AI-driven innovations, such as embedding vectors into streaming data, markedly improves the efficiency and accuracy of data processing. This technology allows UPS to perform real-time analytics, yielding quicker and more accurate decision-making in logistics.
• Upgraded Protection Against Evolving Threats: Striim enables UPS to continuously adapt and enhance its defense models through ongoing analysis of real-time data and dynamic vector generation. This approach significantly strengthens UPS’s capabilities to mitigate evolving threats such as package theft and delivery fraud.

https://vimeo.com/954036063?share=copy

Elevating Logistics Solutions: UPS Capital’s Strategic Partnership with Striim and Google BigQuery

UPS Capital’s adoption of Striim and Google BigQuery represents a proactive and strategic approach to managing the complexities of modern logistics. Through this technological integration, UPS Capital has enhanced its ability to secure packages, optimize delivery routes, and maintain a competitive edge in the logistics industry. The initiative demonstrates how leveraging cutting-edge technology can address the challenges of modern package delivery, ensuring safety and reliability for customers globally.

Discover how Striim on Google Cloud can empower real-time intelligence for AI, just like UPS’s DeliveryDefense Address Confidence. 

Sign up for a free trial today!

View Case Study

Artificial intelligence (AI) uses information to make important choices in different industries. Just like a tall building needs a strong plan and base, successful AI requires good data models. While traditional business intelligence and reporting use cases allow some margin of error, hallucinating AI in customer experiences can be costly for a brand. So how do we create these models for complex algorithms and systems? Organizing, storing, and accessing data is important for AI. It affects how well AI programs and apps work. Incorporating AI into data modeling relies on fundamental techniques and principles that enhance the synergy between data and AI models.

These five key principles will show us how. Data modeling for AI involves making a structured framework that helps AI systems efficiently process, analyze, and understand data to make smart decisions:

The 5 Fundamentals

• Data Cleansing and Validation: Provide data accuracy and consistency by addressing errors, missing values, and inconsistencies. Techniques like outlier detection and imputation help make sure your data is reliable and ready for analysis. 
• Feature Engineering: Craft the right features – the building blocks of your model – by selecting the most relevant and informative ones. Techniques like correlation analysis and feature importance scores guide this crucial step.
• Data Transformation: Prepare your data for analysis by handling scaling issues and addressing skewed distributions. Techniques like normalization and min-max scaling ensure all features contribute equally to model training. Striim’s intuitive platform streamlines data transformation processes, saving you time and resources.
• Model Explainability: Build models that not only give you an answer but also explain how they arrived at it. Techniques like LIME and SHAP shed light on the decision-making process, fostering trust and ethical considerations.
• Scalability and Performance Optimization: Design your model to handle growing datasets and evolving needs. Consider distributed computing frameworks and cloud solutions for efficient processing and future-proofing your AI project.

One excellent source for delving deeper into data modeling is the podcast episode by Joe Reis and Ben Rogojan. Data Modeling With Joe Reis – Understanding What Data Modeling Is And Where It’s Going

Key Principles in Action

• Data Quality: According to a study by Deloitte, following good ways to clean and check data can save money and make information more accurate. This, in turn, leads to enhanced model performance, unlocking the true potential of your AI investments. Additionally, addressing data biases and inconsistencies guarantees ethical considerations are embedded within your AI development process.
• Feature Selection: One advantage of concentrating on important features is that it enhances model efficiency. By using feature engineering tools, you can improve efficiency and accuracy by cutting unnecessary noise. Additionally, having fewer features results in better interpretability, making your AI models more transparent and reliable.
• Normalization: A research paper published by the University of Utrecht explains how normalization improves model convergence, leading to faster and more stable model training. This ultimately translates to reduced bias, as normalization prevents features with larger scales from dominating the learning process.
• Model Interpretability: Building trust with users is crucial for AI adoption. A survey by PwC found that 73% of respondents would be less likely to trust an AI system if they couldn’t understand how it makes decisions. Using methods like LIME and SHAP can help you create trust and transparency in your AI development, promoting ethical thinking and responsible AI practices.
• Scalability: As your AI ambitions grow, your data models need to keep pace. A scalable and flexible architecture provides your models can handle growing data volumes and adapt to evolving needs. This, combined with cloud-based and distributed technologies, future-proofs your AI project, guaranteeing it remains effective and efficient even as your data landscape expands.

Business Value

By following these principles and employing the capabilities, you tap into the full potential of AI efforts and achieve:

• Improved decision-making: Gain deeper insights from data-driven predictions, leading to informed business decisions that drive growth and competitive advantage across various domains.
• Enhanced operational efficiency: Use reliable AI models to save time and work more efficiently. These models can handle tasks, make procedures easier, and improve resource allocation. This frees up human resources to focus on more important activities.
• Increased customer satisfaction: Deliver personalized experiences and tailored offerings based on accurate AI-powered predictions, fostering deeper customer engagement and loyalty.
• Reduced risks and improved compliance: Promote ethical AI development and address potential biases by implementing responsible data modeling techniques, protecting your organization’s reputation, and following changing regulatory standards.

Striim: Your Partner in Effective Data Modeling

Striim’s platform provides the means for integrating and analyzing data in real-time, greatly improving the data modeling process when applying AI. Here are four instances where Striim’s solution can aid in the successful incorporation of AI in data modeling:

• Real-Time Data Integration and Streaming: Striim enables the ongoing collection of data from a diverse range of origins, such as databases, logs, message queues, and cloud platforms. This feature guarantees that the data used for AI modeling remains current, a critical element for applications that rely on real-time analytics or decision-making. By supplying fresh data, AI models can make more precise forecasts and choices based on the most recent information, improving their dependability and efficiency.
• Data Preprocessing and Transformation: Before creating models, data needs to be prepared by cleaning, transforming, or combining it. Striim provides real-time data processing to take care of these tasks as data flows into the system. This makes the modeling process faster and enhances model accuracy by providing that high-quality, properly formatted data is used in AI models.
• Scalability and Efficiency: AI and machine learning models need a scalable infrastructure for efficient data processing and analysis. Striim’s platform can handle large volumes of data in real-time, providing efficient and effective data modeling as data grows.
• Enhanced Decision Making with Real-Time Analytics: Striim’s real-time analytics can inform decision-making with immediate insights and AI integration for adjusting strategies, optimizing operations, and predicting future trends. This allows for a dynamic and responsive approach to data modeling with continuous updates and improvements based on real-time data and analytics.

With Striim’s real-time data integration, processing, and analysis capabilities, we’re basically giving AI a shot of espresso – making it faster, stronger, and more accurate than ever before. Say goodbye to slow and clunky data modeling, and hello to a sleek, efficient, and scalable AI-driven solution.

Ready to unlock the power of AI with effective data modeling?

Get a free demo of Striim today and experience the difference for yourself!