Video Transcript (English)

In today's fast-paced retail and manufacturing worlds, getting inventory right is more crucial, and frankly, tougher than ever.

The biggest challenges for modern inventory systems are about real-time accuracy and scalability. Legacy systems often struggle with stock distribution across many places, different data formats, and slow, batch-based updates. This can lead to big problems like unsold stock and operational bottlenecks.

To tackle these challenges, you need a database and an architecture that can handle data coming in real-time, manage flexible data structures, and easily connect with automation tools.

In this video, we'll dive into how MongoDB Atlas was used to meet these exact technical needs for a company called Thorne & Fable. We'll walk through how they built an inventory system that gave them one clear view of all their stock, could react quickly to demand changes, and helped them make smart, timely decisions.

First off, let me introduce you to Thorne & Fable, a home goods retailer. After a huge holiday season, their old inventory system just couldn't keep up with the complicated task of tracking thousands of items across all their physical stores, online platforms, and a growing network of regional warehouses.

This led to some really frustrating issues. For instance, 'phantom stock' – where products showed up as available online, but were actually nowhere to be found in the warehouse. On top of that, orders were often delayed because data updates were slow and out of sync. And their team was constantly playing catch-up, bogged down by manual checks and outdated batch processes that simply couldn't handle the large volume of transactions.

Lucy, Thorne & Fable’s Supply Chain Director, was put in charge of fixing this. Her team had done everything they could to alleviate the issues – create demand forecasts, make manual stock transfers, even pay for expensive overnight shipping to balance inventory. But nothing worked. The existing setup wasn't designed for the complex, distributed way they needed to manage inventory.

Lucy realized the truth: their system wasn't just a little off, it was fundamentally outdated.

So, her team started looking for a new approach. They needed something that could break down data silos, synchronize inventory across every single location, and react in real-time.

That’s when Sabbir, the team’s engineering lead, suggested a complete re-architecture using MongoDB Atlas. His core idea was to build an event-driven system with one central, real-time data layer. MongoDB's document model could bring all their different inventory data together, and its operational features could handle instant updates and scale easily.

The first step was moving away from their old relational database. In that system, product details like “skews”, variants, locations, suppliers, and transactions were spread out across many different tables. This old setup meant that getting any kind of full inventory picture required complex queries that joined many tables, and running reports often meant slow, resource-heavy batch jobs.

Sabbir then redesigned the data model using MongoDB's document structure. This meant putting all the related product data into a single, complete document. So, each product document now included its variants, stock levels for each location, and all its other details, all in one place. This design drastically cut down on the need for database joins, making data retrieval much faster and simplifying how the application handled inventory queries.

Plus, this new model allowed for instant, atomic updates to product documents as soon as events like online sales happened.

To automate inventory workflows, Sabbir's team implemented MongoDB Atlas Triggers.

They set up a trigger to fire automatically whenever a product document was updated and a specific location's stock level dropped below a set threshold. When this 'low stock' event was detected, the trigger would run a serverless function that automatically created and sent a replenishment order to the right warehouse system.

This event-driven automation replaced all the manual monitoring and ticketing, dramatically cutting down the time it took to move stock.

For real-time updates in the user interface, MongoDB Change Streams was integrated. They set up a Change Stream listener on the inventory collection, which reflected stock changes in real time by providing a continuous, ordered flow of every data change – whether it was an insert, update, or delete. This stream immediately pushed relevant inventory changes to all connected frontend applications.

This meant that sales staff and online customers always saw the most current stock levels. This architectural pattern was incredibly effective at eliminating those 'phantom stock' issues and ensured every department, from sales to fulfillment to customer service, had the exact same, up-to-date view of inventory.

Lucy had one last request: she wanted product search to be much easier, moving beyond dropdown menus.

To achieve this, Sabbir used MongoDB Atlas Search. This allowed the engineering team to build a sophisticated search interface within MongoDB. Atlas Search is powered by Lucene, but doesn’t require complex synchronization across systems.

Staff could now perform full-text searches across product names, descriptions, categories, and even use detailed filters for availability, store location, or delivery window.

The operations team also needed better insights. Before, pulling reports meant waiting for end-of-day exports, or even worse, manually updating spreadsheets. So, for the operations team, MongoDB Atlas Charts was implemented. This gave them a live dashboard directly on their inventory data.

A major architectural benefit here with Charts is MongoDB Atlas's workload isolation capability. This means that when the team runs complex reports or views charts, those analytical queries are automatically sent to separate, dedicated machines. These machines handle the reporting load independently, without slowing down the main database, which is busy processing everyday tasks like sales and stock updates. This ensures that getting real-time insights never impacts the speed or responsiveness of the core inventory system.

Within just six months of the rebuild, Thorne & Fable saw a reduction in their inventory-related operational problems. Customer complaints about out-of-stock items dropped, and employees were able to shift their focus from constantly reconciling inventory discrepancies to more productive tasks. Regional managers now had a level of real-time visibility into inventory that they'd never experienced before.

And as Sabbir put it, “We didn’t just modernize, we future-proofed.”

The final architecture for Thorne & Fable's inventory system leveraged: MongoDB’s flexible document model for consolidated product and inventory data, Atlas Triggers for automated, event-driven replenishment, Change Streams for pushing real-time stock updates to the frontend applications, Atlas Search for efficient, product discovery, Atlas Charts for live operational dashboards with workload isolation.

This architectural design provided a flexible and scalable foundation, allowing Thorne & Fable to rapidly adapt to evolving business requirements in the future.

Looking ahead with MongoDB as their core technology, Lucy and her team are now exploring even more advanced applications. This includes integrating IoT sensors for automated shelf monitoring, using AI for demand forecasting, and even automating returns processing.

While this use case focused on retail inventory, its lessons are broadly applicable across retail, manufacturing, logistics, and many other sectors.

Curious how MongoDB could help your team transform operations? Check out MongoDB's Industry Solutions page to discover many other use cases across industries.