The promise that never quite landed

Open the warehouse. Expose measures and dimensions. Invite everyone to explore. That pitch sounded right for a decade. Yet adoption flattened once the early enthusiasts moved on. Leaders kept asking a simple question that the stack could not answer consistently: “Which number is right?”

Two patterns kept repeating:

Conflicting definitions. Marketing “Qualified Leads” did not match Sales “Qualified Leads”. Finance closed the quarter with a different “Revenue” than the product team’s dashboard.
Choice overload. Dozens of tables, hundreds of fields, and a hunter’s map of joins left business users anxious. You can build anything. You also can build confusion.

"I'm puzzled by the persistently low adoption rates fo BI/analytics tools" Wayne Eckerson

Self‑serve did not fail because users are not capable. It failed because the stack placed the burden of definition and trust on the last mile.

The diagnosis: where self‑serve breaks down

Ambiguous trust. If two dashboards show different values for the same KPI, users stop believing either one. Once trust is broken, usage drops and shadow spreadsheets return.
Surface area problem. Catalogs, glossaries, and semantic descriptions help people find data. They do not stop ten slightly different versions of “Active Users” from being rebuilt across teams.
Help‑desk trap. Facing unclear logic and too many paths, users file tickets. The data team becomes a reactive support queue instead of a partner.

What semantic layers solved, and what still hurts

Semantic layers translate technical models into business names, reduce repeated SQL, and let tools speak a common vocabulary. That shift mattered. You got fewer bespoke queries and cleaner joins. You also got faster onboarding for new tools.

The gap that remains sits at the level you actually manage the business: metrics.

Dimensions and measures give people building blocks. They do not guarantee that “Churn Rate” means the same thing in every deck, dashboard, or AI assistant. Without versioned, tested KPI definitions, ambiguity creeps back in.

Meet the metric layer

A metric layer turns KPIs into first‑class, governed objects. Think of each metric as a small product:

Business‑language definition. A human‑readable name, description, owner, inputs, filters, and grain. For example, “Active Users” might be “distinct count of user_id where last_login is within the past 30 days.”
Version control and tests. Changes are certified, reviewed, and shipped like code. Error notifications catch broken inputs or out‑of‑range values before users see them.
Lineage and explainability. Every metric shows data sources, transformations, and filters so a leader can click into the why behind the number.
Multi‑endpoint delivery. The same definition serves dashboards, BI, product analytics, Slack digests, and AI assistants. One source, many surfaces.

You move complexity upstream, then expose clean, trusted answers downstream. Business users still self‑serve, just within a frame that protects the meaning of the numbers.

How it fits in your stack

Here is the mental model that keeps teams aligned:

Warehouse and transformation (for example, dbt) → Metric layer (metric catalog, ownership, tests) → Consumption tools (PowerMetrics dashboards, BI, product analytics, AI assi

Jonathan Taylor

Guides

Start your analytics journey with instant metrics

How to: Create calculated metrics in PowerMetrics

News

Set goals and hit your business targets in PowerMetrics