I am frequently asked how our data, analytics, and AI team at Chalhoub is organised. While data leaders generally agree on roles such as data engineers or data analysts, what their teams' structure and operating model should look like is less clear. Do you stick all the architects together? Where does machine learning happen? Whose job is it to interface with the business? This is not an academic question as the team's form can directly impact its function.

The standard option is the ‘sorting lego bricks’ approach. You take all the data engineers and make them a team. You take all the transformation talent and stick them in a team. You take all the data quality roles… well you get the idea. This is effectively what I did in my previous role at Dyson. While its simplicity is an advantage, you not only end up with a lot of direct reports, but it can also be difficult for the business to understand what all those different teams do.

Unlike balloons, data teams powered by hot air are not a success. Photo by Aaron Burden.

When I came to Chalhoub, I only had about eight weeks to pull together an initial strategy and budget before they had to be presented to the board. To help land the message, I decided to simplify my proposed team, organising it into three pillars: Data Assets, Data Products, and Analytics Impact. For a retailer like Chalhoub, this structure was intuitive as it mapped to Inventory, Ventures, and Operations. At the same time, it clarified our accountabilities:

Data Assets are the folks tasked with figuring out what data we have, what it means, and whether we can trust it. Additionally, they consider how data should be modeled, e.g. what metrics should be defined and how, based on business needs. Eventually, the output of these architecture, quality, and governance experts should culminate in the creation of a high quality, well-documented machine learning feature store to accelerate model development.

Data Products are the team who focus on developing intellectual property through a combination of data, machine learning, and platform engineering. These can range from simple products like real-time commercial analytics through to complex machine learning to determine optimal customer engagement strategies. At every stage, the product managers in the team should be working with the business to identify and capitalise on opportunities.

Analytics Impact contains a mix of data science, transformation, and operations talent. Their mission is straightforward: to ensure that the business is able to leverage the data assets and products to transforms their operations. Whether conducting analyses, facilitating trainings, or running experiments, every element of their work focuses on tangible business outcomes. Without them, Data Assets and Products would remain isolated intellectual efforts.

Many data efforts get lost at sea without a firm hand on the tiller. Photo by Daniel Kuruvilla.

Defined accountabilities do not preclude shared responsibilities. For example, while accountability for data quality assessments sits with the specialists in the Data Assets pillar, analysts in the Analytics Impact pillar have a critical role to play in validating assets and products with the business. Similarly, data architects will prototype data transformations that are scaled by engineers. Consequently, our product teams at Chalhoub include talent from all pillars.

The upshot is that this model has not only streamlined our team's strategy, operations, and management, but it has also simplified stakeholder communications. Everyone understands how these pillars align with the three goals of an effective data & analytics function: to manage data, create IP, and realise value. In this model, one cannot help but consider the balance between these elements, ensuring a well-rounded team that delivers sustainable value.

– Ryan