This content originally appeared on DEV Community and was authored by Ivica Kolenkaš
“How many white shirts do I need?” is a very simple question to answer for you and me. Answering that same question as a demand planner in a fashion enterprise requires a data driven approach, because the consequences of being wrong are far reaching. To be data driven, you must have actionable data.
The following blog series describes the path that BESTSELLER’s Data & Analytics Platform started 2.5 years ago in an effort to produce actionable data. It will reason about the choices we made, focus on key challenges we faced, and celebrate the big wins that we got from it.
The first step on this path is to understand the business and the platform we’re building for it.
The business
BESTSELLER is a family-owned fashion business that is a home to more than 22000 colleagues across design, logistics and tech. They work for more than 20 brands, including JACK & JONES, ONLY and VERO MODA, across 75 countries.
Each brand in this multi-brand matrix organisation operates with a high degree of independence, which allows them to remain agile in their decision making process. Brand’s identities are different, but their operational practices remain the same. That’s why shared functions, such as the Data & Analytics platform provide the common building blocks for the brands to use.
During daily operations, clothes are sold through multiple sales channels (retail, wholesale, online etc.) which produces large amounts of operational data. It makes every sense to use this data for analytical purposes – to understand the past and predict the future. But with the high degree of independence comes the responsibility to exercise it responsibly. Over the years, several data silos formed, each with its own governance practices, levels of data maturity and ownership structures.
These silos made answering the question “How many white shirts do I need?” very difficult, because the data you need to answer it is scattered, possibly unavailable and under unclear ownership. Data producers and consumers started to become connected in an ever-growing web of point-to-point connections. Even worse, those connections were between clients and databases located on-premise or in the cloud, semi-accessible data stores, semi-structured files and even data stores on personal laptops.
The business is ambitious and we have clear goals:
We want to open one retail store each working day in 2026.
To live up to the expectations of the business, and to answer “How many white shirts do I need?” reliably, across 20 brands and across multiple sales channels, we needed a structured solution that provides actionable data.
The platform
The data & analytics platform (the platform) we’re building has a very clear goal – to enable data and machine learning (ML) engineers to create data products and make them readily available to various parts of the business. These data products can be schemas or views in a database, (semi)structured files unloaded to blob storage, various reports compiled for executives or anything else in between.
Data products are used by various departments in the company to understand the past and predict the future. When used by business analysts and decision makers, these data products help in demand planning, supply chain optimizations, understanding the environmental impact and so on. They augment the decision making process with data.
When describing it to prospective stakeholders, we describe the platform as a highway, which enables you to drive from point A to point B. You could be driving a small electric car or a diesel-powered truck; the highway serves the same purpose. The signage is uniform, the speed limits are known in advance, and the rules apply to every vehicle.
We are building our platform to be:
- The way, but not in the way.
- Flexible, while having general rules.
- In service of those who are using it.
These three tenets shape our vision, decisions and priorities.
If I had to describe the platform in a single sentence, I would say that it is a curated collection of tools, standards and processes to ingest, store, transform and serve data.
The second article in the series (to be published next week) explains the architecture chosen for the platform and what challenges it is meant to solve.
This content originally appeared on DEV Community and was authored by Ivica Kolenkaš