September 13, 2017
| Zoetermeer (NL)
PostNL builds up its team’s data science and machine learning expertise at The Analytics Academy
Last month, The Analytics Academy (TAA) offered a very successful course on data science and machine learning for PostNL employees. This training effort will enable the organization to become more data-driven and future-proof.
PostNL is a mail, parcel and e-commerce corporation with operations in the Netherlands and other European countries. With an eye on the future, the organization wants to apply data science to reduce costs, increase revenue and create new business models based on data products and services. PostNL just launched a department called ACE to build up these capabilities in-house. ACE (short for Analytics Center of Excellence), helps business units improve their processes to increase customer satisfaction. The department’s experts provide data science services to other teams, making use of machine learning and other techniques. Their analyses lead to valuable new insights that contribute to improve operational processes or sales activities.
Mapping PostNL’s training journey
Our collaboration started with an intake scan of PostNL's team, where we tested participants’ level of data science knowledge. The scan enabled us to understand what the focus of the training effort should be and identify the best starting point. Together, we designed, created and organized a customized training program taught by teachers from the University of Amsterdam and ORTEC Data Science. Trainees included members of the ACE team, and data experts from various parts of the organization. Training people outside of ACE was a step forward to stimulate the growth of an analytics community within the entire organization.
Comprehensive data science training
We kicked off the training program with a data exploration, cleaning and visualization session. This was followed by a brainstorm on finding new data science solutions and improving communication between business units and data scientists. Subsequently, we delved into advanced statistics. After covering the essentials, we taught trainees a wide variety of machine learning methods, covering both practical and theoretical aspects. We discussed the following models:
- Decision Trees
- Regression Trees
- Naïve Bayes
- K-means Clustering
- Neural Nets
- Logistic Regression
Besides machine learning models, we discussed meta-methods like ensemble learning, parameter tuning and cross validation.
Testing the skills transfer with a hackathon
At the end of the multi-day course, we organized a hackathon for trainees. During the hackathon, PostNL’s team developed a prototype of a predictive model, putting their new skills into action. The goal was to predict whether a customer of a telephone company was about to switch to a competitor. A jury, which included Marcel Krom (CIO at Post NL), Frank Ferro (IT-director Logistic Services, Post NL), Coen Schut (CFO Logistic Services, Post NL), Marc Salomon (Dean at the University of Amsterdam Business School), Ronald Buitenhek (Center of Excellence Machine Learning at ORTEC), and Marcel Aarts (Director ORTEC Data Science), assessed the results.
We are very proud of this customized training program we prepared with PostNL. It was great to see their newly acquired skills in action during the hackathon!
Interested in building up your team’s data science capabilities? Don’t hesitate to contact Robert Monné
by email, or by phone: +31 6 15 29 13 11 .