Building a Metric System for Sustainable Growth: From Data to Action

01 Build Metric System

First, introduce the concept of a metric system. The core is to find the North Star metric, the single most important indicator that reflects the product’s core value, such as total GMV for a cashback business.

After identifying the North Star metric, decompose it into related factors (e.g., traffic, activation rate, retention) and assign responsibility to relevant teams.

Example: For a cashback business, GMV is split into food‑delivery GMV and e‑commerce GMV, each further broken down into paid users, order value, etc., to pinpoint actionable factors.

Based on this decomposition, design data collection points and use SDKs to report events.

After data collection, build dashboards to monitor metric fluctuations and detect anomalies.

02 Design Strategy Growth Optimization

Introduce the Lift model, a widely used A/B testing framework with six principles: value proposition, relevance, clarity, attention, anxiety, and urgency.

Explain how each principle influences user motivation and conversion, using examples such as coupon incentives and clear call‑to‑actions.

Detail the six principles with visual examples, showing before‑and‑after copy improvements that increase relevance, clarity, attention, anxiety, and urgency.

03 Metric Growth Idea

Return to the earlier case to illustrate growth ideas. For a low delivery page arrival rate (7%), three A/B experiments were run: improving the entry tab visibility, enhancing the landing page with clearer offers, and emphasizing the coupon value.

Results showed significant improvements in arrival rate and user activation.

04 Douyin Group Case Practice

Apply the same methodology to Douyin cases. First, the "Dongchedi" case split video play metrics into play count and per‑user plays, identified bottlenecks in homepage recommendation flow, and improved them via A/B testing, achieving a 300% increase.

Second, address low login rates for older users by simplifying privacy consent prompts, resulting in a 0.5% overall lift (millions of users).

Overall, the process is: build a metric system, find bottlenecks, design growth strategies, iterate with A/B tests, and repeat.