Data Engineering › Data Modeling for Analytics

Conformed Dimension

One shared dimension used by many fact tables so numbers line up.

Data Engineer · Senior · core Data Analyst · Mid-level

Also known as: conformed dimensions, shared dimension, conformed dimension in warehouse, master dimension

A conformed dimension is a dimension that’s shared, identical in meaning and structure, across multiple fact tables (and often multiple data marts). The same dim_customer or dim_date is used by sales, support, returns and marketing, with the same keys, the same attribute names and the same definitions.

                    dim_date (conformed)
                   /        |         \
   fact_sales ────┤   fact_returns ────┤   fact_support_tickets ────┐
                   \        |         /                              │
                    dim_customer (conformed) ◄───────────────────────┘

Why it matters

It’s what lets you combine data from different business processes and get answers that line up. Because both facts point at the same customer dimension, you can ask: “revenue, returns and support tickets per customer segment per month”, by querying each fact separately and joining the results on the shared dimension attributes. This is called drill-across.

Without conformance:

Sales calls a customer’s region “APAC”, and support calls it “Asia Pacific”. The two numbers can’t be put side by side.
Customer IDs differ between systems, so the same person is counted twice.
The “month” in one report starts on the 1st, and in another on the fiscal period boundary.

Everyone ends up arguing about whose numbers are right (metric discrepancies).

What “conformed” requires

The same keys: a shared surrogate key, with consistent mapping from each source’s customer ID (surrogate keys).
The same attribute names and meanings, and the same permitted values.
The same granularity (or a clearly defined subset of attributes shared at a coarser level, often called a rollup dimension).
A single owner and a defined process for changes.

How it’s planned

The bus matrix lists business processes (rows) against shared dimensions (columns). Dimensions used by several processes are the ones to conform first. It guides incremental building of a warehouse: each new process reuses existing conformed dimensions, so the result is integrated, not a set of silos (dimensional modeling).

Practical challenges

Source systems disagree: multiple definitions of customer or product, duplicate records and inconsistent hierarchies. Conforming means deciding and resolving: matching records across systems, choosing the golden version (master data management).
Organizational work: agreeing on definitions across departments is harder than the technology. It needs sponsorship and governance.
Change management: modifying a conformed dimension affects every fact that uses it, so changes must be coordinated and versioned.
Slowly changing attributes must be handled consistently (SCD types).
Role-playing: the same dimension may serve several roles (order date, ship date) in one fact (role-playing dimensions). That’s conformance too.

Date and time dimensions are the easiest to conform, and a good first example (date dimension). The principle also applies outside Kimball-style warehouses, wherever shared reference data and consistent definitions make numbers line up.