Mastering Data Metrics and Tags: Build Powerful Indicator Systems

01. What Are Metrics and Tags? Differences

Metrics are standards used to define, evaluate, and describe specific entities, usually numeric (e.g., new users, cumulative users, user activity rate, monthly revenue, gross margin, net margin). Tags are human‑defined identifiers derived from business scenarios and algorithms, such as customer segment tags (long‑tail, high‑net‑worth) or product risk tags (high, low). Tags combine defined rules with factual data to group entities.

Metrics and tags can be transformed: a metric can be derived from a tag (e.g., migration rate of high‑net‑worth customers), and a tag can be created from a metric (e.g., private‑bank customers defined by AUM ≥ 10 million, where AUM is the metric).

Metric Classification

Metrics are typically divided into three categories:

Atomic Metrics : Basic, non‑dimensional descriptions obtained directly via SQL (e.g., total customers, total projects, total cost).

Derived Metrics : Atomic metrics combined with one or more dimensions (e.g., projects in Guangzhou, projects stopped in Zhanjiang).

Derivative Metrics : Calculations between metrics such as averages or ratios (e.g., average project value, proportion of completed projects, loan‑to‑contract ratio).

Tag Classification

Tags fall into three types:

Factual Tags : Objective, static attributes describing an entity (e.g., whether a component is a procurement item, gender of an employee).

Rule Tags : Processed data combining attributes with logical rules (e.g., overweight cargo, hot‑selling product).

Model Tags : Business‑value‑oriented, predictive labels generated by algorithms (e.g., consumer upgrade potential: high, medium, low).

Metric Usage Scenarios

Metrics are usually broken down into dimensions for monitoring and evaluating business performance (e.g., total customers → mobile‑bank customers, 7‑day incremental customers).

Tag Usage Scenarios

Tags are used for summarizing and characterizing groups, supporting classification‑based strategies such as precise marketing, qualification checks, and personalized product recommendations (e.g., “one‑person‑one‑view” in a banking app).

02. How to Build a Metric System?

Building a metric system combines top‑down and bottom‑up approaches. Start with strategic “north‑star” metrics (e.g., AUM for a bank), decompose them into sub‑metrics aligned with business modules, and also collect frontline metrics from daily operations. Merge both to form a complete hierarchy.

Key steps include business interviews, departmental classification, ensuring metrics are mutually exclusive and collectively exhaustive, and mapping them to the full business process chain to avoid gaps.

Metric definitions must include:

Business Attributes : Unique ID, name, meaning, business scope.

Technical Attributes : System field name, data type, calculation logic.

Management Attributes : Owner department and responsibility for monitoring fluctuations.

03. How to Build a Tag System?

The tag system mirrors the metric system: start from business goals, define tag categories, attributes, and management mechanisms. Clarify which business value layer each tag belongs to, identify missing tags, and supplement accordingly.

04. How to Use Tags and Metrics Effectively?

Common issues include mismatched expected vs. actual tag coverage, duplicate processing, and unclear calculation scopes. Regularly audit tag usage, track popularity, and retire obsolete tags. For metrics, decompose large indicators into sub‑metrics to pinpoint root causes of anomalies (e.g., loan growth slowdown caused by internal approval bottlenecks or external economic factors).

(Source: Architecture Practitioner’s Path)