Building a Data Metrics Center: Essential Skills for Effective Business Analysis

01 Business‑Data Closed Loop

Business and data form a mapping relationship: data is the digital representation of business actions such as user preferences, purchase habits, and daily routines. The more a business is digitized, the richer the data becomes, creating a virtuous cycle where better data further enhances business.

Data → Information → Knowledge

Raw numbers (e.g., a temperature of 39°C) become meaningful only after interpreting the underlying information (fever) and deciding on a knowledge‑based action (seek medical care). Similarly, analysts turn collected business data into insights that guide decisions.

Data Empowering Business

The empowerment process consists of four stages: data performance, business cause, business strategy, and execution method. For example, a fever (data) indicates illness (cause); the doctor prescribes treatment (strategy); nurses administer medication (execution); and continuous temperature monitoring ensures recovery.

Logical and Business Closure

Two inter‑dependent loops ensure effective analysis:

Logical closure : The analytical argument must fully support the conclusion.

Business closure : The strategy must be executed, monitored, and iterated in the real business environment.

Common problems include logical arguments that are sound but not grounded in business reality, and strategies that fail to materialize or receive delayed feedback.

Evaluating Strategy Viability

To judge whether a strategy is “grounded,” follow two steps:

Identify the business assumptions underlying the strategy.

Conduct research to verify those assumptions.

Example: A merchant‑benefit plan may assume merchants understand and care about the plan. If research shows otherwise, the plan must be adjusted before rollout.

Data Analysis and Indicator System

Indicators are the first step in data analysis. They quantify business goals (e.g., registered users, total payment amount, order conversion rate) and are grouped by type: stock, transaction, conversion, ratio, statistical, ranking, etc.

How to Build an Indicator System

Define indicators and assign them to thematic domains . Use the “Warehouse Model Center” as the domain repository.

Separate atomic and derived indicators .

Atomic indicator: a raw statistical field (e.g., count of orders).

Derived indicator: a combination of atomic indicators with dimensions (e.g., daily orders for premium members).

Specify production logic for each indicator . Decompose a derived indicator into its atomic metric, time window, aggregation grain, and dimension constraints.

Standardize indicator naming (concise, uniform format, consistent generation). Example:

Marketplace‑User‑Last7Days‑New‑Suborder‑Daily‑Avg‑PaymentAmount

.

Standardize statistical granularity (units such as “people,” “times,” “transactions” based on the metric’s nature).

Define indicator levels .

Level 1: Core atomic indicators managed by the data platform with a formal development workflow.

Level 2: Derived indicators created by business teams following naming rules.

Roles and Responsibilities

Effective metric governance requires clear boundaries:

Business Product Manager : Aligns product development with business needs.

Data Development Engineer : Processes raw data according to product specifications.

Data Analyst : Builds analytical frameworks, evaluates business health, and guides decisions.

Data Product Manager : Coordinates data engineering, modularizes business data, and productizes analysis frameworks.

Operations : Designs short‑term activities that showcase long‑term product value.

In practice, these roles collaborate to collect data in the warehouse, transform it, analyze business conditions, and deliver insights for product iteration, operational campaigns, and strategic planning.

Conclusion

Constructing a data metrics center involves systematic indicator definition, rigorous governance, and cross‑functional collaboration, enabling data to continuously inform and improve business decisions.