摘要
摘要
近年来，大数据逐渐受到社会和企业的重视，越来越多的公司立足大数据，
进行个性化的精准营销推荐。杭州顺网科技股份有限公司作为中国最大的网吧平
台服务提供商、行业首家上市公司、触达 3亿互联网网民，研究和探索面向用户
的大数据个性化推荐策略意义深远且重大。
基于扎根理论，本研究厘清包括业务价值导向、用户偏好导向、用户感知导
向和市场热度导向四种个性化推荐策略类别；采用专家访谈、层次分析法构建了
个性化推荐策略实施效果评价指标体系，运用评价指标体系对顺网科技个性化推
荐策略实施效果进行评价，发现存在问题主要有：业务价值导向策略推荐准确率
低、投诉量大，用户偏好导向策略预测推荐覆盖率差，用户感知导向策略覆盖率
低、点击率低，市场热度导向策略准确率不稳定、时效性较差。产生这些问题的
主要原因为：业务价值导向策略遵循投放优先原则，用户偏好导向策略受制于标
签滞后性，用户感知导向策略缺少对小客户以及新客户的感知分析，市场热度导
向策略市场热度参考值难以界定。并设计优化方案：业务价值导向策略兼顾商业
投放和用户偏好，用户偏好导向策略增加商业投放曝光机会，用户感知导向策略
运营团队结合研发团队尽可能多开发和提供 SKU，市场热度导向策略充分利用大
数据加以实时或准实时的分析应用。保障措施包括：技术保障、组织管理和系统
维护。
关键词：大数据；个性化推荐策略；实施效果；指标体系
I
Abstract
Abstract
In recent years, big data has gradually been valued by society and enterprises.
More enterprises have made personalized and precise marketing recommendations
based on it. Hangzhou Shunwang Technology Co., Ltd., the first listed company in the
industry is the largest Internet cafe platform service provider in China, which has
reached 300 million Internet users. Its research and exploration on user-oriented big
data personalized recommendation are far-reaching and significant.
This study uses groundedtheory to clarify fourcategories of personalized
recommendation strategies,including business valueorientation, userpreference
orientation, user perception orientationand market popularity orientation. Expert
interviews and AHP areused to construct anevaluation index systemfor the
implementation effect of personalized recommendation strategies which evaluates the
implementationeffect ofShunwangTechnology'spersonalized recommendation
strategy, finding that the main problems are low recommendation accuracy of business
value-oriented strategy,a largenumberof complaints,poor coverageof user
preference-oriented strategy prediction recommendation, low coverage rate of user
perception-oriented strategy, low click rate, unstable accuracy and poor timeliness of
market popularity-oriented strategies. The main reasons for these problems are as
following. The business value-oriented strategy follows the principle of delivery priority.
Theuserpreference-orientedstrategyissubjecttolabellag.Theuser
perception-oriented strategy lacks the perception analysis of small customers and new
customers. For the market popularity-oriented strategy, it is hard to define the market
popularity referencevalue.The optimizationplans areas following.Business
value-oriented strategies take into account commercial placement and user preferences.
User preference-oriented strategy increases commercial placement. The operation team
of user perception-oriented strategies develops and provides as many SKUs as possible
with the R&D team. The data shall be applied in real-time analysis for the market
popularity-oriented
strategy.
Safeguard
measures
include
technical
support,
organizational management and system maintenance.
KeyWords: Big data; Personalized recommendation strategy; Implementation effect;
Indicator system
II