Tamar Alphaidze

2 results

From Chaos to Control: Real-Time Data Analytics for Airlines

Delays are a significant challenge for the airline industry. They disrupt travel plans, erode customer loyalty, and inflict significant financial losses. In an industry built on precision and punctuality, even minor setbacks can have cascading effects. Whether due to adverse weather conditions or unforeseen technical issues, these delays ripple through flight schedules, affecting both passengers and operations managers. While neither group is typically at fault, the ability to quickly reallocate resources and return to normal operations is crucial. To mitigate these disruptions and restore passenger trust, airlines must have the tools and strategies to quickly identify delays and efficiently reallocate resources. This blog explores how a unified platform with real-time data analysis can be a game-changer in this regard especially in saving costs. The high cost of delays Delays from disruptions, like weather events or crew unavailability, pose major challenges for the airline industry. These delays have significant financial impact according to some studies, costing European airlines on average € 4,320 per hour per flight . They also create operational challenges like crew disruptions and reduced airplane availability, leading to further delays, which is known in the industry as delay propagation. To address these challenges, airlines have traditionally focused on optimizing their pre-flight planning processes. However, while planning is crucial, effective recovery strategies are equally essential for minimizing the impact of disruptions. Unfortunately, many airlines have underinvested in recovery systems, leaving them ill-prepared to respond to unexpected events. The consequences of this imbalance include: Delay propagation: Initial delays can cascade, causing widespread schedule disruptions. Financial and operational damage: Increased costs and inefficiencies strain airline resources. Customer dissatisfaction: Poor disruption management leads to negative passenger experiences. The power of real-time data analysis In response to the significant challenges posed by flight delays, a real-time unified platform offers a powerful solution designed to enhance how airlines manage disruptions. Event-driven architectural approach The diagram below showcases an event-driven architecture that can be used to build a robust and decoupled platform that supports real-time data flow between microservices. In an event-driven architecture, services or components communicate by producing and consuming events, which is why this architecture relies on Pub/Sub (messaging middleware) to manage data flows. Moreover, MongoDB’s flexible document model and ability to handle high volumes of data make it ideal for event-driven systems. Combining these features with PubSub’s, this approach proves to offer a powerful solution for modern applications that require scalability, flexibility, and real-time processing. Figure 1: Application architecture In this architecture, the blue line in the diagram shows the operational data flow. The data simulation is triggered by the application’s front-end and is initialized in the FastAPI microservice. The microservice, in turn, starts publishing airplane sensor data to the custom Pub/Sub topics, which forwards these data to the rest of the architecture components, such as cloud functions, for data transformation and processing. The data is processed in each microservice, including the creation of analytical data, as shown by the green lines in the diagram. Afterward, data is introduced in MongoDB and fetched from the application to provide the user with organized, up-to-date information regarding each flight. This leads to more precise and detailed analysis of real-time data for flight operations managers. As a result, new and improved opportunities for resource reallocation can be explored, helping to minimize delays and reduce associated costs for the company. Microservice overview As mentioned earlier, the primary goal is to create an event-driven, decoupled architecture founded on MongoDB and Google Cloud services integrations. The following components contribute to this: FastAPI: Serves as the main data source, generating data for analytical insights, predictions, and simulation. Telemetry data: Pulls and transforms operational data published in the PubSub topic in real-time, storing it in a MongoDB time series collection for aggregation and optimization. Application data: Subscribed to a different PubSub topic, this service acknowledges static operational data, including initial route, recalculated route, and disruption status. Therefore, this service will only be triggered provided any of the previous fields are altered. Finally, this data is updated in its MongoDB collection accordingly. Vertex AI integration—analytical data flow: A cloud function triggered by PubSub messages that executes data transformations and forwards data to the Vertex AI deployed machine learning (ML) model. Predictions are then stored in MongoDB. MongoDB: A flexible, scalable, and real-time data solution Building a unified real-time platform for the airline industry requires efficient management of massive, diverse datasets. From aircraft sensor data to flight cost calculations, data processing and management are central to operations. To meet these demands, the platform needs a flexible data platform capable of handling multiple data types and integrating with various systems. This enables airlines to extract valuable insights from their data and develop features that improve operations and the passenger experience. Real-time data processing is a must-have feature. This allows airlines to receive immediate alerts about delays, minimizing disruptions and ensuring smooth operations. In fast-paced airport environments, where every minute counts, real-time data processing is indispensable. For example, integrating MongoDB with Google Cloud's Vertex AI allows for the real-time processing and storage of airplane sensor data, transforming it into actionable insights. Business benefits This solution provides real-time access to critical flight data, enabling efficient cost management and operational planning. Immediate access to this information allows flight operation managers to plan ahead, reallocate existing resources, or even initiate recovery procedures in order to diminish the consequences of the identified delay. Moreover, its ML model customization ensures adaptability to various use cases. Regarding the platform’s long-term sustainability, it has been purposely designed to integrate highly reliable and scalable products in order to excel in three key standards: Scalability The platform’s compatibility with both horizontal and vertical scaling is clearly demonstrated by its integral design. The decoupled architecture illustrates how this solution is divided into different components—and therefore instances—that work together as a cohesive whole. Vertical scalability can be achieved by simply increasing the computing power allocated to the designed Vertex AI model, if needed. Availability The decoupled architecture exemplifies the central importance of availability in any project’s design. Using different tracks to introduce operational and analytical data into the database allows us to handle any issues in a way that remains unnoticeable to end users. Latency Optimizing the connections between components and integrations within the product is key to achieving the desired results. Using PubSub as our asynchronous messaging service helps minimize unnecessary delays and avoid holding resources needlessly. Get started! To sum up, this blog has explored how MongoDB can be integrated into an airline flight management system, offering significant benefits in terms of cost savings and enhanced customer experience. Check out our AI resource page to learn more about building AI-powered apps with MongoDB, and try out the demo yourself via this repo . To learn more and get started with MongoDB Vector Search, visit our Vector Search Quick Start page .

October 15, 2024

How to Enhance Inventory Management with Real-Time Data Strategies

In the competitive retail landscape, having the right stock in the right place at the right time is crucial. However, the retail industry faces significant challenges in achieving this goal. In 2022, unsold stock in the US surged by a staggering $78 billion, reaching approximately $740 billion—a shocking 12 percent increase . Without a single view of inventory, retailers struggle to compete with new market disruptors offering customers omnichannel experiences. Retailers who get stock management right can move to distributed supply chains, leveraging stock across online and in-store platforms to distribute inventory quickly and react to shifting buying patterns. With effective access to the data, retailers speed up workforce efficiency and allow for automation. In this blog, we will explore how inventory management affects customer experiences, effective stock management for accurate demand forecasting, and workforce productivity. Building a single view of inventory to enhance customer experience Modern retail consumers expect seamless omnichannel experiences, like the ability to view product availability online and pick it up at a nearby store the next day. They will gravitate toward retailers that prioritize their need for convenience and speed. The difficulty in delivering these features often stems from the lack of a centralized inventory hub, i.e. operating with separate inventories for online and in-store. Combining data from diverse sources, including vendor solutions, RDBMS databases, and files, becomes a complex task that hampers the ability to achieve an accurate real-time view of stock availability. It also extends the time to market for new features, requiring redundant and customized development efforts across different channels. This lack of adaptability impacts the retailer's ability to offer customer-centric features, putting them at a disadvantage compared to their competitors. To track inventory in real-time and improve visibility and consistency across multiple channels and locations, MongoDB’s document data model is a powerful choice. Using the document model, data types can be combined easily, making it more flexible for handling diverse product data. Its intuitive design enables developers to iterate on the data model at the same pace as the rest of the code base, without downtime for schema changes. This agility accelerates the implementation of new features and functionalities that can be built on top of a single view of inventory, like real-time stock availability, and buying online and picking up in-store the next day. Figure 1: Enabling buy online and pick up in-store through single-view inventory By leveraging a single view of inventory, retailers can accelerate the development of superior customer experiences, securing a competitive edge in the retail industry. Effective stock management with real-time analytics Now that the retailer can see and understand inventory levels across their organization in one place, they can begin to manage stock more effectively. This enables retailers to move to a more complex distributed supply chain and activate the use of real-time analytics or AI. In a traditional retailer without a centralized inventory management system, the complexity of mixing stock between channels was too difficult in a segmented data landscape, leading to waste through dead stock in stores while others or online channels have an insufficient supply of the same item. With a single view of inventory, items can be moved around in a way that makes sense for the business. Online orders destined for in-store pick-up might be packed using in-store items. Dead stock on a shelf might be available online. Stores can move stock between themselves in an intelligent manner. The added complexity does come with more complex decision-making. It's vital to be able to ask difficult questions about the inventory management system and get answers in real-time. Rather than move data to a different analytical platform and get answers a day later, retailers are looking to do real-time analysis to make important stock allocation decisions in real-time. Next, retailers tackle demand forecasting and bring intelligence into stock allocation. This is where a translytical data platform comes in. Its distributed architecture means analytical workloads can run on a real-time analytics node. This approach eliminates the need for additional systems such as separate analytics platforms and reduces the lag associated with transferring data. The aggregation framework, MongoDB’s advanced processing pipeline can then be used to ask complex analytical questions and get results back to the user in real-time. For example, retailers can easily see which products are the most popular or the most likely to run out of stock soon or understand when a product rapidly sells out in one store if this is a trend or tied to a specific event like a sports game. This insight can guide smart decisions on redistributing products to get them in front of the customer who is most likely to buy. Figure 2: Inventory real-time analytics This architecture could also be leveraged to feed AI or machine learning models. The more complex the supply chain becomes, the more retailers are turning to cutting-edge technology to gain further insight. Demand forecasting is a great use case for AI as there can be a vast amount of possible factors and results. With MongoDB, retailers are integrating AI systems so they can access real-time data, enhancing their accuracy and responsiveness. This synergy enables businesses to streamline their supply chains. Boost workforce efficiency through an event-driven solution A successful inventory management strategy also contributes to improving workforce efficiency. The lack of real-time updates brings on inefficient inventory tracking procedures that result in errors, such as excess or unavailable goods, and hinder customer orders, leading to dissatisfaction among staff and customers alike. As the business grows and sales volume increases, the ability to process large amounts of real-time data becomes increasingly important. A future-proof, scalable, and flexible architecture supporting the tools that empower your workforce, can make a difference when retailers face a peak in demand or decide to expand their business. The central question retailers face is, "How can businesses enhance workforce efficiency in their inventory operations? The key lies in using event-driven architectures for managing inventory systems. MongoDB is a great fit for this approach, offering features like Change Streams , Triggers , and the Kafka Connector . Take for example the scenario seen in Figure 3; a customer purchases a t-shirt in-store. The Point of Sale device then instantly updates the product stock. If stock runs low, this change is instantly sent to the store manager app through Change Streams to alert the store manager. To automate the re-ordering process, MongoDB Triggers can be set up to trigger a function that would perform complex actions in response to the event, like automatically reordering products. Figure 3: Event-driven architecture for inventory management Today, when an influencer mentions a particular item, it can fly off the shelves at an unforeseen pace. Thanks to automation enabled by event-driven architectures, such situations become opportunities, not challenges. As soon as that item goes unexpectedly out of stock, the system triggers an automatic reorder, ensuring that your shelves are replenished in real-time. This rapid response eliminates the need for manual intervention, freeing up your store manager to focus on more value-add activities. Instead of spending hours every day reordering items, they can now dive into more engaging tasks, like interacting with customers, providing personalized recommendations, and exploring innovative stock decisions. This isn't just a theoretical advantage. A prime example comes from MongoDB’s work with 7-Eleven . By implementing a custom inventory management app, 7-Eleven streamlined its operations across 10,000 stores in the U.S. and Canada. With event-driven functionality, 7-Eleven store employees can now seamlessly manage transactions, sales, and inventory through mobile devices, eradicating the need for manual updates and improving overall workforce efficiency. Closing the loop for a future-proof inventory management strategy Effective inventory management strategies are vital in the evolving retail landscape. By providing a consistent single-view inventory, retailers can enhance customer experiences and gain a competitive edge. With efficient stock management capabilities, they can optimize their inventory levels, reducing costs and improving profitability. And by embracing event-driven solutions, retailers can boost workforce efficiency, enabling data-driven decision-making and streamlining processes through automation. If you want to get hands-on, follow our step-by-step tutorial on how to Build an Inventory Management System using MongoDB Atlas . Access our GitHub repo for code samples, video guide, and more!

August 23, 2023