bonprix

Sascha Netuschil

Introduction

In the dynamic landscape of AI development, organisations are constantly seeking effective ways to streamline their processes and maximise efficiency. Product organisation has long been a popular approach in the software development industry. But this organisational principle has also found its way into the realm of data science and AI development. When it comes to establishing in-house AI development teams, product organisation has emerged as the most preferred method.

As an internationally successful fashion company – with online shops in more than 25 countries and reaching over 16 million active customers we at bonprix have been developing in-house AI solutions for over seven years. Thus, the question about the optimal form of organisation has been in our focus regularly.

The adoption of product organisation brings numerous advantages, including efficient resource allocation, accelerated development times, and a culture of agility and continuous improvement in product development. Recognising these benefits, we at bonprix decided to re-organise our AI development teams into product teams. Over the years, we have gained valuable experience with this organisational setup, further solidifying our belief in its effectiveness.

Guiding Principles of AI Product Organization at bonprix

At bonprix, we have established guiding principles for the product organisation of our AI product teams. Our teams take end-to-end responsibility for their AI products and have all the necessary skills, resources, and technological prerequisites to operate independently. As we believe in the power of self-organisation and self-responsibility, our product teams have the autonomy to define their vision, goals, and priorities. They are accountable for the development and operation of their products, as well as determining the most effective path to achieve their goals. This fosters a sense of ownership and dedication among our teams, driving innovation and continuous improvement.

An AI product at bonprix is defined as an AI system and its output. This can range from personalised product recommendations for every user to performance predictions for every article. To illustrate the responsibilities of our product teams, let me give you an example. If the digital product is the predicted customer lifetime value (CLV) of all customers, the product team is accountable for the accurate prediction and the timely and reliable delivery of the prediction values into the destination system. Additionally, it is responsible for documenting the data and providing support for its usage. However, it is important to note that the product team is not responsible for the raw data used by the AI model or the subsequent processing and usage of the product.

"As we believe in the power of self-organisation and self-responsibility, our product teams have the autonomy to define their vision, goals, and priorities"

Product organisation is closely intertwined with agile product development methodologies. We leverage various agile methods that align with our needs and discard those that do not work for us. This flexibility allows us to adapt and evolve our processes while staying true to the core principles of product organisation.

Setup and Roles of AI Product Teams

AI product teams at bonprix consist of four key roles, each with its specific responsibilities:

• The product owner takes charge of product management, including defining the vision, roadmap, and focus. This person also handles requirement management, prioritization, and stakeholder management, and acts as the spokesperson and first point of contact for the team.

• The data scientists are responsible for AI model development and methodology, including feature engineering and data analytics.

• Machine learning (ML) engineers take charge of data pipelines and the productive process. They are responsible for monitoring and alerting, ensuring that the AI models and systems are running smoothly and efficiently.

• The agile master is responsible for overseeing the agile process within the team. This role ensures that the team adheres to agile principles and practices, facilitating smooth collaboration and efficient delivery of results.

Typically, bonprix teams have one product owner at least two data scientists and two ML engineers. The aim is to maintain a balanced ratio of 1:1 between data scientists and data engineers. The role of the agile master is shared among the teams, allowing for cross-functional collaboration and knowledge exchange.

Conclusion

At bonprix, the adoption of product organisation for our AI development teams has proven to be a true game-changer. It empowers our teams, fosters innovation, and ensures end-to-end responsibility for our digital products. By aligning with agile product development principles and defining our digital products, we have created an environment that promotes efficiency, collaboration, and continuous improvement. As we continue to leverage product organisation, we are confident in our ability to drive impactful AI solutions and deliver exceptional customer experiences.