How to Build an Effective Data Metric System for Data Governance

1. Overview of Data Metrics

Data metrics differ from traditional statistical indicators; they are aggregated results derived from data analysis that quantify business units, making goals describable, measurable, and decomposable. Common metrics include PV, UV, and business‑specific measures such as total playback time (minutes).

Metrics consist of three components: dimension (the perspective), aggregation method (how data is summed), and measure (the unit, e.g., minutes).

2. Principles of Building a Metric System

2.1 Focus on Priorities – Avoid merely listing metrics; assign priority and clarify which metrics are examined first.

2.2 Define Clear Objectives – Metrics should be tied to concrete business problems rather than generic dimensions.

2.3 Relevance Over Completeness – The most valuable metric system is the one that best fits the specific business context, not the one that tries to cover every possible indicator.

3. Designing the Metric System

The design process is divided into four stages:

Identify Requirements – Determine the product lifecycle stage (early, middle, late) and align metrics with strategic goals, business‑driven needs, or optimization loops.

Define Primary Metrics – Choose top‑level metrics that reflect overall product health (e.g., acquisition, activation, retention, revenue, referral, recall – the AARRRR model).

Derive Secondary Metrics – Break down primary metrics into actionable sub‑metrics linked to specific strategies (e.g., revenue → ad revenue, in‑app purchase revenue).

Derive Tertiary Metrics – Further decompose secondary metrics to pinpoint responsible teams or processes (e.g., in‑app purchase → browse product → add to cart → submit order → payment success).

Example: The metric "total playback time" is measured by the dimension (time period), aggregation (sum of minutes), and measure (minutes).

4. Implementing the Metric System (Event Tracking)

Implementation focuses on event tracking rather than primary metrics. Key steps include:

Prepare a tracking specification document covering workflow, naming conventions, and requirement details.

Obtain product or activity prototypes to define page and element names.

Define event names using a structured format (behavior_object_result_type), e.g., click_purchase_success.

Identify event dimensions using the new 4W1H framework (Who, When, What, Where, Why, How) and capture causal relationships.

Determine reporting timing (display, click, or API response) for each event.

Produce a data requirement document summarizing all tracked events.

Enter metrics into a metric dictionary, categorizing them by business domain for easy lookup.

Choosing between self‑built data portals and third‑party tools (e.g., Sensors, GrowingIO, ZhugeIO) involves trade‑offs: self‑development offers flexibility but higher effort; third‑party services provide quick setup but may limit custom calculations.

5. Methodologies and Experience

Effective metric system construction relies on solid thinking models:

5W2H – Who, What, When, Where, Why, How, How much – to ensure comprehensive analysis.

Logical Tree & MECE – Decompose complex problems into mutually exclusive, collectively exhaustive sub‑issues.

Business Canvas – Map business model elements to uncover hidden data needs.

Additional frameworks include the "First Critical Metric" concept (focus on one key metric per stage) and the AARRRR lifecycle model for user‑centric metrics.

6. Value of a Metric System

Three main benefits:

Standardized Business Quantification – Enables consistent measurement and comparison over time (e.g., detecting a drop in monthly net profit).

Increased Efficiency – Reduces ad‑hoc data requests by covering most analysis needs within the system.

Rapid Problem Diagnosis – Allows back‑tracking from high‑level changes to root causes via linked primary, secondary, and tertiary metrics.

All benefits presuppose reliable data quality; otherwise, even a perfect metric system yields meaningless insights.

Images