The organisation, workflow, and technology basis of the core platform was agreed with our client within the first three weeks. The product team and engineering team met and began their collaboration, building a trusting and transparent relationship. The combined team was small, at only 18 people, and the no-blame, collaborative, can-do culture we put in place from the outset was vital to the success of the project.
By week 21…
In the next 21 weeks, we delivered a total of 2,389 production releases, 18 microservices, a customer user interface, an iPhone app, a data lake ingesting 14 data sources, along with two data labs and the first data science algorithm. By this stage the platform’s throughput was 10 million transactions per day.
By week 60…
By week 60, the project had grown: 5,742 production releases, 78 microservices, new customer and management interfaces, Android and iOS apps, 32 sources for the data lake, and four machine learning algorithms in production. A key measure of the platform’s reliability is its uptime: 100%.
Realising the value of machine learning
To fully realise the value of machine learning in marketing, we had to understand more about what the likes and dislikes of its customers to predict what they might buy. In our machine learning project we created algorithms to work out the value of individual shoppers, based on frequency, recency, spend, product range, loyalty and length of relationship, expressing this value as an affinity score. The next step was to determine the best channels to communicate with these shoppers though hyper-personalised offers to compound loyalty and drive sales.
The loyalty card programme has over 15 million customers, and in order for the machine learning project to fulfill its potential, we needed to break down the barriers between data scientists, engineers, and marketers. The data scientists train algorithms drawing on a data lake combining dozens of sources, and these algorithms are then used to deliver an automated marketing campaign. A digital dashboard produces reports for data scientists, engineers, marketers, and senior management at the client.
