How Data Assets, Tag Systems, Metric Frameworks, and User Profiles Interconnect

Enterprises building data capabilities often encounter four concepts: data assets, tag system, metric system, and user profiles.

Is customer spending a data asset or a metric?

Is a "high‑value customer" a tag or a user‑profile segment?

Why build a tag system when a metric system already exists?

Does a user profile finish by simply displaying data on one page?

These questions arise because the four concepts operate at different layers of data processing and business application.

The data warehouse is the foundational layer that aggregates dispersed source data, cleanses, integrates, and models it around entities such as customers, products, orders, and channels, thus providing a trustworthy base for downstream metric calculation, tag generation, and user‑profile construction.

1. The essence: not four independent systems

Data assets, metric system, tag system, and user profiles form a processing chain rather than parallel systems:

Data assets supply raw material.

Metric system quantifies facts.

Tag system judges features.

User profiles organize the results on specific objects.

Separating these layers leads to disconnected data catalogs, untraceable tag rules, and stale user‑profile data.

2. Data Asset Management

A data asset must be identifiable, interpretable, traceable, and compliant:

Identify where each data field originates and which business process creates it.

Provide clear business meaning for each field (e.g., what "customer status" really means).

Track the data flow through synchronization, cleaning, association, and calculation steps.

Define ownership, access permissions, usage scope, and sensitivity level.

Only when these conditions are met does the data become a usable asset rather than a simple inventory.

3. Tag System

Tags summarize business object features and can be classified as:

Fact tags directly derived from business data (e.g., region, registration channel).

Rule tags calculated from conditions (e.g., customers with > 5,000 CNY spend in the last 30 days).

Model tags produced by predictive algorithms (e.g., churn probability).

A production‑ready tag must specify:

The target object.

The calculation rule.

The statistical time window.

Update frequency.

The business scenario that consumes the tag.

For example, a "silent customer" could be defined as no login for 60 days or no transaction for 90 days, each serving different business needs.

4. Metric System

Metrics are quantified results of business facts. A complete metric definition includes:

Statistical object and scope.

Time grain.

Data source.

Handling of abnormal data.

Responsible department.

Policy for recalculating historical data after rule changes.

Example – three possible definitions of "repurchase rate":

Customers with ≥ 2 purchases ÷ all purchasing customers.

Current‑period repurchasers ÷ previous‑period purchasers.

First‑time buyers who repurchase within a defined period ÷ all first‑time buyers.

Although all compute a repurchase rate, each measures a different business aspect and must not be mixed.

Metrics also have hierarchical levels:

Atomic metrics directly based on raw facts.

Derived metrics that add time, range, or dimension constraints.

Composite metrics that combine multiple metrics.

Clear metric hierarchy prevents duplicate calculations and inconsistent results across reports.

5. User Profile Construction

User profiles are not merely a 360° page; they are a data model that supports specific decision scenarios. They consist of four layers:

Identity layer : unified customer identifiers (ID, phone, member number, device, etc.).

Fact layer : behavioral records such as registration, browsing, ordering, payment, refund, and service interactions.

Metric & Tag layer : calculated values (e.g., annual spend, purchase frequency) and classification tags (e.g., high‑value, churn risk).

Action layer : business actions triggered by the profile (e.g., coupon issuance, manual follow‑up, risk review).

A profile without actionable steps is only a static display; only profiles that drive operations create value.

6. End‑to‑End Business Chain

Using a churn‑prediction scenario, the four concepts link as follows:

Gather data assets: consolidate customer info, login logs, orders, refunds, and service records, ensuring consistent identifiers.

Calculate metrics: recent purchase interval, purchase frequency in the last 90 days, spend change rate, refund rate, complaint count.

Generate tags: classify customers as normal, declining activity, or high‑risk churn.

Build the churn‑risk profile by combining identity, factual behavior, metric values, and risk tags.

Trigger actions: high‑value high‑risk customers enter a manual recovery list; silent customers receive automated outreach; customers with unresolved complaints are excluded from marketing pushes.

In this chain, data assets provide trustworthy sources, the tag system translates behavior into comparable values, the metric system quantifies facts, the user profile organizes everything around the individual, and business rules decide the final action.

Thus, data assets, metric system, tag system, and user profiles solve four distinct layers of problems; only when they are integrated into business processes does data become a true asset.