Chow Sang Sang: Real-Time DaaS & Unified Inventory

From Silo-ed Systems to Real-Time DaaS — How Chow Sang Sang Unified Data Across Four Regions and Six Brands

Sep 26,2025

Imagine a shopper walking into a boutique in Hong Kong, while another browses the online store in Shanghai. Both are looking for the same limited-edition necklace. In the past, those two moments might have triggered two separate systems, with no guarantee the stock view was aligned. Today, thanks to a unified real-time data service, the answer is consistent everywhere: “Yes, it’s available — and reserved for you.”

Background & Challenges

Chow Sang Sang (CSS) is a heritage jewelry retailer with over a thousand stores and six brands across Mainland China, Hong Kong, Macau, and Taiwan. Decades of growth left the company with a tangle of business systems, and silo-ed data: more than a dozen disparate ERP, POS, and WMS systems running independently across different regions and brands.

The result was fragmented product data, such as product information and product inventory. The operational systems are similar in nature, however each system had its own set of business logic customized for the local market, making it difficult to deliver a seamless omnichannel experience. For associates, it meant uncertainty when promising stocks. For ecommerce, it meant inconsistencies in product attributes. For IT, it meant endless one-off integrations and long lead times just to launch a new digital service.

Our challenge wasn’t the lack of data — it was that the data lived in silos. Four regions, six brands, more than ten systems. We needed one truth for everyone, everywhere.

The TapData Solution

To overcome this complexity, CSS turned to TapData Live Data Platform to implement a real time DaaS (Data-as-a-Service) data infrastructure. The aim is clear: 1) to provide a consolidated product inventory across the group companies, 2) to significantly shorten the development cylce needed for delivering API services to the marketing and mobile team.

Here’s how it worked.

TapData ingests real-time changes from multiple Oracle ERPs, POS systems, and WMS platforms using built-in high performance CDC pipelines. All inserts / updates /delets are captured as they happen from Oracle’s redo log. These change events unified into MongoDB, where data is standardized into master and application models. From there, incremental materialized views provide low-latency access, and reusable APIs publish trusted domains — Products, Inventory, Orders, Discounts — for downstream systems.

Capture changes at the source Using log-based CDC, TapData continuously monitors Oracle redo log. It then parses database change events like inserts, updates, and deletes as they happen, from multiple Oracle ERP, POS, and WMS systems. Effectively, TapData turned these monitored tables into multiple change streams, and ready for next step processing
Consolidate multiple sources into one single collection The change events captured from source are handed to in-memory data processing pipeline for necessary transformations and standardization. Then these data records were written into centralized data store (a distributed MongoDB database cluster). In CSS’s scenarios, multiple tables from different data sources were written into same collection to form a global view of product inventory
Turn into JSON data model In addition to simple consolidation, with TapData’s Incremental Materialized View feature, we can join multiple table streams and output the result into a materialized view stored in MongoDB, as a rich JSON data model. This API ready JSON data model, or called IMV in TapData, is continously updated by any changes occurred in the source tables. The latency is typically 1~3 seconds.
Expose reusable APIs as DaaS These Trusted domain models — Products, Inventory, Orders — are published as versioned REST APIs. These APIs replace fragile, one-off pipelines with governed, reusable services that every channel can consume.
Ensure scale and availability Deployed in a horizontally scalable architecture, the solution guarantees high availability as well as high performance access across four regions.

For developers, this replaced brittle, point-to-point pipelines with governed data services. For business teams, it meant every channel — ecommerce, boutique POS, membership, after-sales, and analytics — could rely on a single, consistent, always-fresh view of the truth.

For the first time, our teams in Mainland China, Hong Kong, Macau, and Taiwan are working from the same source of truth. It has completely changed how we think about launching new digital services.

Results & Business Impact

With TapData, Chow Sang Sang unified product and inventory data across four regions, six brands, and more than ten independent systems into one real-time source of truth.

All-channel consistency. Associates and customers now see the same stock and product details, whether online, offline, or through clienteling apps.
Faster delivery. New data services that once took weeks to develop are now delivered in days, accelerating digital initiatives.
Confidence and trust. Sales and service teams no longer hesitate before confirming availability. “When we tell a customer an item is in stock, we know it’s accurate — everywhere,” said a senior operations lead.
Future-ready foundation. The ODH + DaaS architecture has created reusable, governed data assets that power today’s experiences and allow rapid rollout of tomorrow’s.

It’s not just faster integration. It’s the confidence that when a customer asks about a product, every channel gives the same answer, in real time.

Why it matters

For luxury retail, the brand promise extends beyond the jewelry itself — it’s about trust in every interaction. By centralizing data into an Operational Data Hub and serving it as Data-as-a-Service, Chow Sang Sang turned fragmented systems into live, reusable data products. The outcome: consistent, real-time information that empowers associates, delights customers, and scales across regions and brands.

The best change is the quiet one. We spend time with customers, not with spreadsheets.

Operational Data Hub Implementation Pitfalls — and How to Avoid Them

Sep 26,2025

Batch Is Broken: Time to Think Incremental

Sep 03,2025