What is data modelling?

Data modelling is the process of structuring and relating your data so that it accurately represents your business and can be analysed reliably. It defines the tables, the relationships between them and the agreed meaning of each measure, turning raw data from systems such as SQL Server, Microsoft Fabric and Power BI into a trusted foundation for reporting, self-service analytics and AI.

What is the difference between a star schema and a snowflake schema?

A star schema places a central fact table of measurable events around simple dimension tables, giving fast, easy-to-understand queries. A snowflake schema goes further by splitting those dimensions into related sub-tables to reduce duplication. Star schemas are usually preferred for Power BI and analytics because they perform better and are simpler to navigate, while snowflake designs can suit highly normalised warehouses.

What is a semantic model in Power BI?

A semantic model in Power BI, previously called a dataset, is the layer that holds your tables, relationships, calculations and business definitions. It sits between your raw data and your reports, so that measures such as revenue or active customers are defined once and used consistently across every dashboard, which is what makes trustworthy self-service reporting possible.

Should I model my data in SQL, Microsoft Fabric or Power BI?

In practice you model across all three, with each playing a role. SQL Server and the Microsoft Fabric Warehouse or Lakehouse are where you shape and prepare clean, well-structured data, while Power BI is where you build the semantic model that reports run on. The right balance depends on your data volumes, your tools and your team, which is something PTR can assess with you.

What is Direct Lake in Microsoft Fabric?

Direct Lake is a storage mode for Power BI semantic models in Microsoft Fabric that queries data directly from OneLake without importing or refreshing it. It combines the fast performance usually associated with import mode and the up-to-date data usually associated with DirectQuery, making it well suited to large models that need to stay both quick and current.

What is the difference between Kimball and Inmon data modelling?

Kimball and Inmon are two established approaches to data warehouse design. The Kimball method builds dimensional star schemas focused on business processes and is quick to deliver usable reporting. The Inmon method first builds a normalised, enterprise-wide warehouse and derives data marts from it. Many modern Microsoft Fabric and Power BI solutions lean towards the Kimball dimensional approach for analytics.

Will data modelling fix my slow Power BI dashboards?

Very often, yes. Slow dashboards are frequently caused by poor underlying models rather than by Power BI itself, including overly wide tables, inefficient relationships and calculations doing too much work at query time. Redesigning around a clean star schema, well-written DAX and an appropriate storage mode such as Direct Lake typically delivers a marked improvement in speed and reliability.

Data Modelling

What's holding your reporting back?

If your data is technically available but still hard to trust, the model underneath is usually the cause. The symptoms tend to look familiar. Reports disagree with one another, dashboards slow down or time out as the data grows and people end up stitching exports back together in Excel to get the number they actually need.

Underneath that, the same metric is often defined differently by every team, there is no agreed definition of something as basic as a customer or revenue and there is no safe way to let people explore the data for themselves. None of these are reporting problems in isolation, they are signs that your data has ineffective modelling.

"Same question, two dashboards, two answers."

Core modelling concepts

Good business intelligence rests on proven modelling foundations and we design around them rather than reinventing the wheel. At the centre of all is the star schema, where measurable facts sit surrounded by descriptive dimensions, an approach that consistently outperforms a single wide table for both speed and clarity. We build these schemas using Kimball dimensional modelling, working from a bus matrix of conformed dimensions and giving careful thought to the level of detail each fact table records.

From there, the detail matters. Slowly changing dimensions let you track history correctly, so a customer who moves region last year does not quietly rewrite last year's figures. Decisions about normalisation and denormalisation, including the trade-offs of a snowflake design, are made deliberately rather than by accident. Above all of this sits the semantic layer, the place where measures, hierarchies and agreed business definitions live, so that everyone draws on the same meaning.

Modelling in practice

Designing the star schema

We start by fixing the grain of each fact table and building conformed dimensions around it, so that the same metrics line up cleanly across every report rather than drifting apart from one dashboard to the next.

The semantic model and DAX

One governed set of measures feeds every visual, giving the business a single source of truth for the number. When a definition changes, it changes in one place and flows everywhere consistently.

What you get

The outcome is a documented, governed model that your whole business can trust, supporting dashboards that are faster and more reliable. Because everyone draws on one agreed set of metrics, self-service business intelligence becomes something you can enable safely rather than fear. Just as importantly, a well-modelled foundation is what makes AI genuinely useful, since they can only be as trustworthy as the data they sit on.

Data Modelling Consultancy Services

Apply meaning and context to your data

What's holding your reporting back?

"Same question, two dashboards, two answers."

Core modelling concepts

Modelling in practice

Designing the star schema

The semantic model and DAX

What you get

"One model. One truth. Every report agrees."

Share This Page

Articles you may be interested in

Frequently Asked Questions