Building Internal Capabilities to Develop Data Science Projects

This guide will take you through the following steps:

1. Problem: How can analysts make use of Data Science?
2. Solution: Getting up to speed in Data Science for Marketing
3. Applications: How to apply Data Science in Marketing
4. Case Study: EDF Energy’s Data Science training and business outcomes

How can analysts make use of Data Science?

It’s important to note that data science capacity does not need to be built as a specific business unit within an organisation. No matter what sector, businesses can benefit from bringing data science into different teams/divisions. In order to provide existing domain experts, such as business analysts, with the knowledge and capabilities to develop new products, enhance productivity and operational efficiency, and perform advanced analytics that leads to better insights and decisions.

In the Marketing sector, Data Science and Machine Learning techniques can help predict customer acquisition, retention and profitability to improve business outcomes. There are a plethora of Data Science tools and techniques available to analysts. Therefore, having the right skill sets on your team, and ensuring analysts understand when and how to apply these techniques to generate business value is essential.

Utilising the skills picked up in this training we have been able to market to our customers and prospects more effectively, using less resources to acquire and retain customers. We’re now moving forward and looking at using some of the methods we’ve picked up as a result of this training to fix some larger business problems.

Matt Wilson, Senior Manager, EDF Energy

Getting up to speed in Data Science for Marketing

The in-person training involved three stages. “The idea was to get people familiar with Python through the Introduction to Python Course,” said Matt Wilson, Senior Manager, EDF Energy. “Then push forward to help individuals gain an understanding of some of the popular Data Science and Machine Learning methods in Python, which they could then apply to business use cases upon returning to work.”

A guide to getting started:

1. Gain familiarity with Python 3 and related libraries for data analysis including:

• Numpy and Pandas for data manipulation
• scikit-learn for Machine Learning
• matplotlib and seaborn for visualisation

2. Upskill in Core Data Science; Essential techniques to learn include:

Data Science Essentials: Data Exploration and Pre-processing

• Pre-processing: scaling, dealing with missing values and outliers
• Feature engineering techniques
• Exploratory Data Analysis using Pandas
• Principal Component Analysis using scikit-learn

Unsupervised Learning

• k-means clustering
• Hierarchical cluster analysis

Supervised Learning

• K-Nearest-Neighbours
• Decision Trees
• Random Forests and Gradient Boosting
• Support Vector Machines
• Logistic Regression

Machine Learning

• Optimisation and Regularisation: implementing gradient descent, regularisation
• Model evaluation: tuning hyperparameters, preventing overfitting and data leakage
• Model interpretability: using LIME, SHAP

3. Link Data Science to your business outcomes

We worked with the team at Cambridge Spark to define a package that would suit our needs specifically — It was absolutely relevant to us.

We’ve actually re-used a lot of the code put together in the training exercises to deploy production models — the workbooks completed as part of the training are referred to by our data science team frequently.

- Matt Wilson, Senior Manager, EDF Energy

Common Applications of Data Science in Marketing

1. Predictive and Prescriptive Customer Analytics
   Using predictive modeling and Decision Optimisation to predict campaign outcomes.
2. Forecasting
   Using Regression analysis and Decision Trees to predict customer conversion, retention and lifetime value based on historical data and customer segmentation.
3. Recommendations
   Building Recommender systems to provide personalised customer-specific product offerings by clustering groups of similar customers based on their historic transactions, interactions and “taste dimensions”.

EDF Energy’s Data Science training and business outcomes

By the end of the year the objective was to have a team of five trained data scientists ready to work on any complex business problems,” said Matt Wilson, Senior Manager at EDF Energy. “Having attended a couple of Cambridge Spark events in the past, I saw the training as the perfect foundation to get the prospective data scientists started.

We’re 5 months on from the training now, the training was very useful for us to build capability quickly, we used it to quickly deploy a series of classification models for marketing optimisation (predicting campaign outcome). We’re still using them now, however, they have been vastly improved on, tuned and fully automated.

The training delivered was engaging and interactive. The trainer was attentive to individual needs and the sessions ran at a pace where no-one was left behind,” said Matt. “I’ve attended many data science courses in the past, many of which are either too focused on the coding side, or too focused on the mathematics. I found the course the perfect balance between the two aspects and therefore a great level for someone with extensive experience as an analyst (using more traditional tools, i.e. not Python) looking to move into Data Science.