Case study: Production-ready code for Data Science at the BBC

Data-driven approach at BBC

The BBC is the world’s most renowned and largest national broadcasting organisation.

Serving television, news, sport, and radio to an international audience of 376 million, they’ve helped define global media for nearly a century.

They take a data-driven approach to understand how people use their services, tracking clicks and content consumption to find out:

• The increase of how often people use the BBC's services, and how long they use them for
• Why some users watch less frequently, helping customise a better experience for them

Developing a bespoke training course

The BBC’s Marketing and Audience team come from backgrounds in machine learning and statistical analysis. Their analytical and exploratory software was already hugely useful in looking at consumer behavioural and audience data.

Their challenge was to encourage and adopt good software development practices to help long term code maintenance. This would allow them to increase productivity, fix bugs more efficiently and reduce the risk of project failure. 

Junior and senior members of the team attended a bespoke Cambridge Spark training session which fused elements from two of our courses:

• Python and Pandas for Analysts
• Data Science in Production with Python

Over the course of the training, we covered how to build efficiency into every step of a project’s life cycle, including:

• The fundamentals of development using Python
• Structuring, configuring and deploying code
• Going from Python Notebook to packaging
• Productionising the model
• Monitoring and debugging code
• Software Engineering principles including testing, code quality, and design patterns

Result: production-ready approach

With practical examples and our experts answering questions as we went, the team could contextualise their skills within a broader framework.

By the end, they knew how to transform their approach and write production-ready code that’s maintainable and in-line with industry practice whilst using K.A.T.E.® to benchmark their code.

The Marketing and Audience team has embedded a new standard for deploying their code which follows modern practices.