Artificial Intelligence 12 min read

OPPO’s Unified Modeling and Smart Power Strategies for App Distribution and User Value

The article details OPPO’s approach to balancing cost reduction and user value in app distribution through unified cross‑scenario modeling, sparse‑data solutions, oCPX advertising optimization, and a multi‑stage smart‑power system that improves efficiency, scalability, and revenue while preserving user experience.

DataFunTalk
DataFunTalk
DataFunTalk
OPPO’s Unified Modeling and Smart Power Strategies for App Distribution and User Value

In the current industry climate, cost reduction and efficiency are essential, but sacrificing user value can jeopardize long‑term competitiveness; OPPO emphasizes that both must be pursued together.

Expert Introduction: Lai Hongke, General Manager of OPPO Internet Application R&D Platform and Search Algorithm Department, has over 14 years of experience in internet advertising R&D, leading the development of OPPO’s ad playback system, search engine, and software store.

OPPO faces a broad range of scenarios in app distribution, from finance to gaming, making recommendation systems more challenging than typical e‑commerce contexts due to data sparsity and long conversion chains.

To address sparse data across dozens of distribution scenarios, OPPO built a full‑scene unified modeling platform that shares global features and samples, expanding KV storage from 2 TB to 30 TB and feature count from 100 million to over 10 billion, thereby improving generalization across scenes.

The platform incorporates MMOE (Multi‑Gate Mixture of Experts) to share embeddings between shallow and deep conversion goals, dramatically reducing the number of CVR models while handling sparse deep‑conversion data such as game payments.

OPPO also introduced the oCPX capability, allowing advertisers to optimize for specific conversion events (view, download, registration, payment) with automated bidding based on predicted conversion rates, improving cost‑effectiveness and user experience.

Smart‑power system evolution:

V1.0: Dynamic throttling and user segmentation to protect high‑value traffic, achieving a 15 % increase in traffic support and revenue.

V2.0: “Fake‑traffic” detection and VIP user identification, boosting traffic support by 20 % and further increasing revenue.

V3.0: Prioritizing the top 20 % of traffic that generates 80 % of value, allocating additional compute resources to VIP users.

These steps illustrate how OPPO integrates algorithmic innovation, engineering design, and business insight to achieve cost reduction without compromising user value.

Artificial IntelligenceData ModelingRecommendation systemsOPPOApp DistributionSmart Power
DataFunTalk
Written by

DataFunTalk

Dedicated to sharing and discussing big data and AI technology applications, aiming to empower a million data scientists. Regularly hosts live tech talks and curates articles on big data, recommendation/search algorithms, advertising algorithms, NLP, intelligent risk control, autonomous driving, and machine learning/deep learning.

0 followers
Reader feedback

How this landed with the community

login Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.