Unified Data Platforms: How UMENG+ Redefines Big Data Strategy

At the beginning of this year Oracle released a big‑data trend report noting that more enterprises are combining user analysis and enterprise applications with big data, moving from AI‑supported apps to data‑stream clients and pursuing personalized experiences.

Chinese big‑data firms are now focusing on building one‑to‑one experiences by collecting and linking user data across multiple channels, shifting business from reactive to proactive and predictive models.

1. Why Call It “Omni‑Domain Data”?

Omni‑domain data blurs the line between online and offline sources, aggregating data from PC, mobile, and physical locations to create a unified view that can drive value.

According to UMENG+ COO Ye Qian, data useful in one scenario often proves valuable in another, prompting enterprises to expand data collection across channels despite the challenge of integrating fragmented datasets.

Key challenges include multi‑channel data collection, cross‑screen data integration, and processing unstructured data.

Historically, companies only tracked basic metrics like retention and daily active users; today they demand granular insights such as age, preferences, frequency, and geography, moving from simple statistics to intelligent analysis and deep mining.

UMENG+ illustrates its approach with the U‑DIP (UMENG+ Data Intelligence Platform) model.

Data sources span PC, mobile, and offline channels, each providing distinct logs that are stored, feature‑extracted, and linked via a universal ID repository.

UMENG+ also builds knowledge bases from web, social, and app‑store data, feeding machine‑learning models to generate demographic and interest tags, and to identify risky or fraudulent devices.

Professor Liu Dehuan of Peking University warns that many current tags are pseudo‑labels, suggesting future paths: deeper human‑centric analysis and object‑centric detection based on user cognition.

2. Aligning Technology with Business Growth

As business deepens, technology must evolve in three areas: data‑collection front‑end, backend computation, and optimization of data‑mining results.

The collection layer has progressed from simple protocols to sophisticated anti‑fraud mechanisms such as virtual‑machine detection.

Backend processing grew from modest workloads to thousands of nodes in Hadoop clusters, later migrating to Alibaba Cloud’s StreamCompute and MaxCompute platforms, with extensive optimizations for large‑scale, high‑throughput tasks.

Early reliance on open‑source stacks (Hadoop, Storm, Kafka, Spark, Elasticsearch) gave way to Alibaba Cloud solutions for stability and scalability.

Mining results now inform system improvements, enabling detection of abnormal devices, traffic, and fraudulent activities, especially in e‑commerce, payments, O2O, and advertising.

During last year’s Double‑11 event, UMENG+ filtered over 34 billion ad impressions, identified 70 million anomalous clicks, and saved advertisers more than 30 million RMB.

3. Unifying the Tech Stack Across Multiple Business Lines

In 2017 UMENG+ planned three business expansions: internet app data, new‑retail data, and advertising‑marketing data, accompanied by a unified technology stack.

Four pillars of unification: a single programming language (Java), a common framework (Spring), standardized components, and a consolidated underlying platform (Alibaba Cloud MaxCompute).

Standardizing languages and frameworks reduces learning curves, improves stability, lowers operational costs, and can shrink hardware expenses.

Adopting a unified stack also requires cultural change, as engineers must move out of their comfort zones and embrace new tools.

4. Delivering Real Value to Customers with Data

Future big‑data trends focus on three points: tighter integration with vertical industries, increased data openness and exchange, and deeper AI‑driven predictive capabilities.

Addressing data‑security extraction challenges remains critical.

UMENG+ aims to help customers overcome data‑utilization difficulties, emphasizing a “customer‑first” mindset and a craft‑spirit culture that values persistence, focus, and deep technical digging.