How Alibaba Turned a Mall into a Fully Digital IoT Ecosystem

Building Digitalization – IoT and Edge Computing

Qinchengli, a large shopping mall next to Alibaba's Xixi campus, opened in 2018 as a fully digitalized mall. From the planning stage it was designed to be "online" by integrating all building subsystems through the self‑developed IBOS platform.

Device Data Dictionary

Building equipment is diverse (HVAC, water, fire, security, power, etc.). To digitize the building, a unified data dictionary was created, encoding over 100 device types across six major systems, defining a complete data model.

Niagara – The IoT Solution

Niagara, a Java‑based open IoT platform, provides edge terminals (JACE) with multiple I/O ports, built‑in drivers for common building systems, plugin‑based driver management, and supports standards such as Haystack, LonWorks, and BACnet, solving heterogeneous subsystem integration.

Niagara addresses two pain points: (1) integration of heterogeneous subsystems, shielding developers from protocol work; (2) standardizing raw device data into Haystack‑formatted messages for IBOS processing.

IBOS Edge Computing

People, devices, and spaces are the three fundamental models. IB‑Connector ingests device data and dynamically maps it to these models based on the data dictionary. The IB rule engine performs real‑time, event‑driven calculations on model instances. The platform also offers flexible components for custom model creation and cloud data delivery.

Data Center Architecture

The IB data center aims to unify and process all online and offline building data. It provides PaaS data services and algorithm services for downstream business decisions. The architecture follows a four‑layer stack:

Data sources – various online/offline data from IBOS, IB applications, Group DT (OneData, A+), and external partners.

Data modeling and computation – data ingestion, cleaning, domain‑specific modeling, and real‑time/ETL processing.

Data service middle‑platform – OLAP analysis and open data services for business domains.

Applications – data‑driven use cases such as leasing, site selection, operation analysis, consumer insights, traffic flow, financial modeling, etc.

Data Modeling

Using Kimball dimensional modeling, the team built a logical model covering all business systems (traffic, parking, POS, CRM, multi‑screen, leasing, property, energy management). The model reduces data redundancy, offers clear structure, supports OLAP analysis, and provides extensibility for new analytical needs.

Data Service Platform

The platform delivers flexible data interfaces via graphical or SQL‑based tools, enabling upper‑level applications to customize their data needs without relying on data engineers. Benefits include unified data standards, easier quality monitoring, reduced data‑market size, and robust API management, monitoring, flow control, and automated testing.

Monitoring includes QPS, response time, error rates, and alerting.

Core Business Scenarios

Traffic

Traffic systems act as the "PV" and "UV" of offline malls. By defining activity models, the system can compute daily, weekly, and monthly active visitors and members, and analyze demographics for resource allocation and performance evaluation.

Transaction

Transaction data is heterogeneous across merchants. Solutions include direct TP integration for Alibaba merchants, deployment of Koubei POS for most retailers, and a sales management system for remaining merchants, with data fed into the warehouse.

Membership

The membership system provides real‑time points, multi‑benefit integration (parking, coupons, etc.), facial‑recognition interactions, and links OneID members with Taobao members, enriching user profiles with offline and online data.

Data Visualization

Standard internal visualization tools (e.g., YouZhu, DataV) are used, and plans are underway to visualize building data using 2D/3D maps, CAD/BIM models.

Special Events

During the first Double‑11, real‑time mall data was displayed on the Group's media screen.

Operations Monitoring

In collaboration with the Cloud Intelligence team, a monitoring model was built on the "XuMiShan" situational platform.