Big Data Insights from the 2015 Internet+ Summit: Advertising, Finance & Security

1. Data Monetization and Trading – History and Future (Liu Peng, 360 Commercial Products Chief Architect)

Liu Peng distinguished true big‑data applications from non‑big‑data ones by focusing on three dimensions: core transaction data (behavior), full‑scale data processing (personal credit, advertising, recommendation), and automated applications such as targeted ads and customer relationship management. He illustrated how advertising remains the most mature big‑data domain, showing revenue increments derived from data‑driven advertising and explaining the shift toward programmatic and native ads on mobile platforms.

Key takeaway: data value can be quantified (e.g., 6000+6000‑10000 = 2000) and the most effective monetization comes from personalized ad delivery versus static impressions.

2. Data‑Driven Operations Creating Business Value (Zhang Ximeng, GrowingIO Founder & Former LinkedIn Analytics Director)

Zhang described how LinkedIn grew its annual revenue from $100 million to $5 billion by leveraging data‑driven user acquisition, achieving a user‑cost half of the industry average, and using real‑time KPI tracking (600 KPIs) to accelerate product testing and conversion optimization.

He emphasized the importance of full‑cycle data analysis: from user attributes to behavior to social signals, enabling precise segmentation and recommendation, and highlighted the growing adoption of data‑driven decision making in both foreign and domestic internet companies.

3. Lean Application Performance Management in the Big‑Data Era (Liao Xiongjie, Vice President of Technology, Tingyun)

Liao stressed that monitoring must keep pace with rapid product iteration. He introduced a lean performance‑management workflow that injects monitoring code automatically (e.g., via Java‑agent or JVM‑level agents) to avoid manual instrumentation overhead.

Example: to evaluate the runtime efficiency of a function xxoo, the team uses automatic code injection rather than hand‑written probes, ensuring developers are not burdened while still collecting precise latency metrics.

4. Data Asset Management in the Big‑Data Age (Cheng Yongxin, Executive Vice President, Xinju Network)

Cheng presented a five‑layer “five‑star” model covering data architecture, governance, sharing, internal asset appreciation, and external scenario monetization. He argued that a robust data architecture drives enterprise‑level maturity and that establishing a data‑asset management committee is essential for overseeing the entire data lifecycle.

Data governance must involve both IT and business units to define standards, manage metadata, and ensure data quality, while data sharing enables rapid modeling and visualization across core business scenarios.

5. Behavioral Prediction Models for Finance (Liu Zhijun, CDO, Capital One Former Director)

Liu explained that the finance team experiments with clustering, prediction, and classification algorithms, finding prediction to be the most valuable. He highlighted challenges such as sample bias, missing values, and the high cost of obtaining reliable credit data.

Model evaluation focuses on relevance and stability, with classic ranking algorithms illustrated in the accompanying diagrams.

6. Agile Data Operations for Internet Finance (Wang Tong, Vice President, Beijing Yonghong Business Intelligence)

Wang advocated “same‑day demand, same‑day data” with high‑performance, self‑service platforms. Early‑stage analysis focuses on the full user lifecycle to boost traffic and conversion, while mid‑stage analysis adds financial and thematic insights.

Exploratory BI is essential because internet‑finance teams often lack sufficient technical resources to meet growing data‑service demands.

7. Security Defense Platform for JD Finance (Liu Minghao, Senior Security Expert, JD Finance)

Liu distinguished technical security (e.g., XSS, privilege escalation) from business security (e.g., fake accounts, fraud, activity cheating). He introduced a risk‑map workflow: tolerance → warning → intervention, and demonstrated how IP and region anomalies can trigger alerts.

8. Big‑Data Platform and Risk Control at MaDai Finance (Wang Tianqing, Chief Architect)

Wang outlined the typical data‑flow architecture used across Chinese internet firms, emphasizing the need for multi‑dimensional data to increase value density. He discussed credit risk categories (credit, information, operation, fraud) and the use of social data and topic modeling to assess creditworthiness.

9. Big‑Data Cloud Practices at Yirong (Zheng Yun, R&D Director, Yirong Big‑Data Innovation Center)

Zheng described Yirong’s massive data ingestion pipeline (one new customer per minute, 20 million loans per hour) and the construction of a user‑behavior collection system feeding a knowledge graph. He highlighted the importance of social‑data dimensions and topic‑model‑based credit scoring.

Overall, the summit underscored that big data is transitioning from raw volume to refined assets, driving advertising efficiency, financial risk assessment, operational agility, and robust security across industries.