How JD’s PLUS Membership Used Data and Algorithms to Drive Growth

Background

As the era of traffic bonuses fades, the internet industry shifts from “traffic is king” to “users are king,” making the retention of existing users a critical business challenge. Paid membership models, such as JD’s PLUS membership launched in 2015, emerged as a successful way to provide exclusive services, improve user experience, and generate revenue at lower acquisition cost.

Technical Growth and Growth Hacking

To move from coarse, labor‑intensive growth tactics to fine‑grained, data‑driven growth, JD’s Algorithmic Intelligence Team began a closed‑loop exploration in Q4 2019, leveraging data, algorithms, and experimentation. The growth‑hacking methodology emphasizes using data and algorithmic personalization to intervene at every stage of the user lifecycle.

Data Construction

A foundational data system was built to support growth analysis, covering three dimensions:

User dimension: reconstructed membership status tables, separating full‑snapshot (DA) and incremental (DT) tables to improve calculation efficiency.

Product dimension: transformed PLUS‑exclusive price product tables into DA status tables and created basic data for rights acquisition and usage.

Traffic dimension: built key‑path data for membership activation to enable full‑link conversion analysis.

All tables are stored in Hive, serving as a pre‑condition for downstream analysis and product development.

Data Analysis Framework

The analysis follows four layers:

Descriptive analysis : explains what happened using statistical summaries, trend analysis, and the “people‑goods‑place” model to assess current membership health.

Diagnostic analysis : investigates why events occurred through qualitative causal analysis (questionnaire surveys) and quantitative attribution analysis (external and internal attribution methods such as last‑click, first‑click, linear, Markov‑chain, Shapley value, correlation, root‑cause, and DuPont decomposition).

Predictive analysis : forecasts future outcomes (e.g., membership expiration) and predicts user actions using machine‑learning models.

Prescriptive analysis : provides actionable recommendations, currently manual but moving toward automation.

Data Productization

Using JD’s internal EasyBI platform, the team built an offline membership growth dashboard and a real‑time marketing analysis module. After three weeks of development, the first version of the PLUS membership growth dashboard was launched, followed by iterative enhancements and an automated report that surfaces key KPI fluctuations, channel‑level attribution, and renewal forecasts.

Algorithmic Foundations

Algorithms are divided into two categories:

Strategy algorithms : rule‑based pipelines (e.g., recall → ranking) that are interpretable and low‑cost, used for rapid MVP validation such as the “Save Money Guide” campaign.

Model algorithms : machine‑learning models (supervised and unsupervised) for tasks like clustering, association rule mining, classification, and regression, applied to predict membership activation and click‑through rates.

Growth Experiment: Rights Recommendation

The experiment followed a standard growth‑loop:

Idea : personalize rights push to increase perceived value and renewal rates.

Data validation : qualitative surveys revealed low awareness of rights; quantitative correlation showed strong link between rights receipt and renewal.

Hypothesis : improving rights perception will boost renewal.

Prioritization : high‑impact, data‑validated, and algorithmically feasible.

Design : push messages with layered targeting (user state, rights type, copy) and grouped into control, random, and model arms; a supervised learning model predicts click‑through probability.

Implementation : configuration‑driven XML for copy and links, enabling rapid iteration.

Effect : significant lifts in rights penetration, activation numbers, and renewal rates; results were embedded in the dashboard for quick issue detection.

Summary and Outlook

The algorithmic intelligence team demonstrated that a data‑centric, algorithm‑enabled approach can substantially improve membership growth. Future work aims to integrate root‑cause analysis and automated reporting into an augmented analytics platform, explore natural‑language generation for insights, and build a unified product matrix that combines data analysis, algorithms, and personalized messaging.

Gartner’s 2019 “Augmented Analytics” trend underscores the direction toward automated data preparation, insight discovery, and sharing for broader business users.

Key Visuals