Customer Story

BESTSELLER turns massive datasets into faster, more reliable BI insights with Tabular Editor 3

bestseller-customer-success-story

One simple solution

to handle large and complex datasets

Faster delivery

of more reliable and up-to-date BI insigts

Master Model Pattern

used efficiently to simplify the semantic model setup

BESTSELLER

Industry

Retail

Company Size

+25,000

Location

Global, 90 countries
BESTSELLER is an international fashion company that produces, sells, and distributes clothing and accessories. Its portfolio of more than 20 brands, including JACK & JONES, ONLY, and VERO MODA, is sold in over 90 countries worldwide. The business operates across three channels: retail, wholesale, and e-commerce. BESTSELLER operates approximately 3,100 branded retail stores in 47 countries, supplies more than 16,500 multi-brand and department stores worldwide through its wholesale business, and offers products from all 20+ of its brands through its e-commerce channels.

BESTSELLER’s BI and planning functions are organized into specialized teams covering wholesale, retail, e-commerce, planning and the underlying data foundation. Common to all teams is the goal of delivering well-structured, accessible data that enables users to quickly find the insights they need, whether into current business operations or historical data.

Henrik Hornshøj (Solution Architect), Jonas Ebsen (Senior Data Analyst), and Christian Priess (Data Engineer) work in BESTSELLER’s BI and planning teams. In this case, they share how Tabular Editor has helped BESTSELLER transform and simplify its semantic models while enabling faster delivery of more reliable, up-to-date BI insights.

BESTSELLER handles enormous amounts of data

BESTSELLER’s BI planning team forecasts how much clothing needs to be produced and purchased and runs replenishment for the stores. These are driven by a background engine that uses sales and stock data from each store to continuously calculate the optimal quantity of each clothing type per location, then automatically generates sales orders to keep stock levels sufficient.

For this process to work, the BI team depends on accurate, always up-to-date data to monitor how sales are progressing. Every day, BESTSELLER receives sales and stock data from approximately 8,000 locations, a mix of its own stores and partner stores. As a result, millions of rows are added to the existing semantic model every day.

One of our tables alone holds around 8 billion rows. Getting that to perform is quite a task.
Henrik Hornshøj, Solution Architect, BESTSELLER
The level of detail adds to the challenge. Specifically for retail, BESTSELLER’s users need to work at the EAN level, where each style is broken down into attributes such as size, length, color, and print. To keep this manageable, the team has split the product dimension into several smaller ones.
Instead of one big dimension, we’ve split it into separate dimensions for color, size, length, and more, and that helps quite a bit. But our users need to see exactly which size and color is sold in a given country under a given area manager – and that’s heavy, because a tabular model is better at aggregating than at showing very low granularity.
Jonas Ebsen, Senior Data Analyst, BESTSELLER

On top of that, store performance is compared like-for-like. A layer of measures on top of the existing measures dynamically filters which stores were also open the year before and indexes this year’s sales against last year’s. This kind of business logic is essential, but it also makes the model heavier.

At this scale, handling the data as efficiently and optimally as possible is essential to ensure the latest insights are always available to the organization and its decision-makers.

Tabular Editor 3 helps BESTSELLER overcome challenges in its semantic model setup

Working with brand-specific models is part of the solution

To handle these enormous amounts of data more efficiently, BESTSELLER has built brand-specific models using Tabular Editor’s Master Model Pattern. With this setup, BESTSELLER maintains one master semantic model, from which perspectives are deployed as separate models using scripts. As a result, the team can work with smaller models, which have made a noticeable difference in their everyday work.
We have one master model, and from that we use scripts to deploy perspectives as separate brand models. It’s actually a pretty smart setup. It lets us keep the models smaller, even though we still need the combined model for anything that goes across brands.
Henrik Hornshøj, Solution Architect, BESTSELLER

Partitioning larges tables

With millions of rows added every day, full refresh is not an option. The team uses partitioning strategies to refresh only what is needed and finds this intuitive and easy to set up in Tabular Editor.

Working with Perspectives in Tabular Editor 3

The brand-specific setup relies heavily on perspectives, and this is where Tabular Editor 3 makes the difference. Team members who start out in Tabular Editor 2 quickly find it too limited when working with perspectives as BESTSELLER does and must upgrade to Tabular Editor 3 to work with the setup.

Choosing Tabular Editor was a no-brainer

When BESTSELLER started using Tabular Editor in 2016, there was no real alternative. The company’s models were so large that other tools could not support the setup without significant load times, making Tabular Editor the natural choice.
Other tools either couldn’t handle our large tabular models at all or made working with them far too troublesome and time-consuming. Tabular Editor was the obvious choice, and we’re still just as happy with that decision today.
Henrik Hornshøj, Solution Architect, BESTSELLER
The BI team highlights these features as the most transformative for their BI work:
  • Master Model Pattern: Scripts deploy brand-specific models from the master model, reducing model size and improving performance.
  • Best Practice Analyzer (BPA): The built-in BPA rules in Tabular Editor 3 help the team identify optimization opportunities in their models.
  • IntelliSense: Supports the team’s work with BESTSELLER’s complex DAX, helping them keep track of it and write DAX measures correctly.
  • Workspace mode: Lets the team work with data and metadata simultaneously, which they highlight as a great advantage.
  • Vertipaq Analyzer: Provides summary statistics on the model's contents and recommendations to optimize and reduce the dataset size.

Tabular Editor 3 enables more efficient ways of working

With Tabular Editor 3, BESTSELLER has significantly improved how efficiently it handles Power BI models of this size and volume.
As soon as you have a Power BI model of a certain size, working in Tabular Editor 3 is the logical choice. It’s easier, faster, and smoother when you’re working with data volumes like ours.
Jonas Ebsen, Senior Data Analyst, BESTSELLER
Despite the volume of data, the team describes Tabular Editor as a simple, stress-free tool for working with large, complex datasets. And with Snowflake as BESTSELLER’s data platform, the team also benefits from new Snowflake features being implemented faster in Tabular Editor 3 than in any other solution.
I’ve worked exclusively with Tabular Editor 3 for many years, and I wouldn’t choose to work with anything else. You can work with your data in real time without waiting for anything to load. It simply works.
Christian Priess, Data Engineer, BESTSELLER

Looking ahead

The BI team sees great potential in scripting to simplify both their work and their datasets. They already use AI to help write advanced C# scripts, allowing them to complete tasks such as creating temporary test models more quickly.

Going forward, the team plans to explore the MCP Server in Tabular Editor 3. After seeing a demo, they immediately saw great potential and quickly came up with a range of use cases they want to explore.