Cover Image
Author's Image
Sebastian Karrer
Best practices for the implementation of data science projects
Transformation processes are a complex subject and digital transformation is no exception. However, there are proven best practices that companies can use to set up an analytics program right from the ground up and gain a sustainable competitive advantage.

According to a McKinsey study, companies that use analytics for their business are 50% more likely to achieve higher growth rates than their competitors. Data-driven B2B companies get a 5% higher return on sales than their less digitized competition. The consultancy’s take is that many companies have great potential by utilizing the information that is buried in their databases, by learning more about customers, their own processes and the market environment. Digital champions across a wide range of verticals prove that by using these insights to gain competitive advantages a company can potentially change its entire industry.

Corporate transformation is often a complex and multi-dimensional process. In the case of digital transformation, this is further complicated by the fact that both the data analytics method and the management of data are new challenges for many organizations. Executives must therefore have a very good plan tailored to its own company on how the transformation project can master these additional complexities.

The difference between digital champions and companies that are less digitized and use analytics little or not is not always because the digital champions invest more in the area. Rather, the success of the digital champions comes from the fact that they approach digitization issues and analytics more systematically and strategically.

There are generally three important framework conditions for successful analytics initiatives that every company should keep in mind:

  1. The program aims to improve those decisions and processes that make up the core of the company’s business
  2. All employees buy in to using data-driven methods with the personal goal to achieve better results
  3. Each project is not only to yield operational improvements, but also to give the company knowledge and new ideas

Successful analytics programs do not start with a huge financial investment, but they start with developing right approach.

Digital champions first define a vision at the management level of how the analytics initiative will enhance the core business of the company. This results in the data and analytics strategy and a first collection of use cases and the setup of the first data team. All use cases are based on the business, are consistent with each other and contribute to the overarching goal.

We looked at these strategic considerations in another article. The following is about operational best practices from digital champions, which every company should adopt when implementing its analytics projects.

Implement data strategies efficiently: the agile, iterative approach in the analytics project

When implementing their strategy, analytics champions proceed step by step: With a divide & conquer approach, they break down large use cases into individual projects. Each project can be implemented quickly, delivers a measurable result and is tangible for everyone involved. In this way, analytics champions reach the proof-of-concept of the entire use case more quickly and they continuously improve their approach by using the knowledge gained from each individual project. And they focus on the use cases that have a big impact on the business.

At the project level, Analytics Champions use agile and iterative processes. This minimizes the complexity, because the focus is always on the concrete, known next steps. At Erium we use the following scheme:

Analytics Project Scheme

  1. Idea Validation The innovation idea is examined using available data. The result is a qualified assessment of the feasibility of the project, the data quality and instructions for the next steps
  2. Prototyping The idea is being developed further, for example using an improved dataset and an adapted model. A prototype achieves the goals formulated at the beginning
  3. Implementation The idea is put into practice by integrating the further optimized model into the target process or by incorporating the knowledge gained into practice.

This scheme maps the knowledge gained in the project and means that the correct questions are processed for each status. Iteration one is comparable to fundamental research, iterations two and three are similar to the design and implementation phases of a product development. In each iteration cycle, the project goes through a structured process, with the team naturally shifting the focus as necessary. In iteration one, the team will primarily deal with the data, iteration two is about optimizing the model and in iteration three an algorithm must be developed that is ready for deployment.

Within each iteration, the team jumps back to a previous step if necessary. If, for example, a calculated correlation cannot be further developed into a model, the team goes back to the data preparation, e.g. to check whether there are any influencing variables that are not included in the data set as a separate category.

Best practices for the development of database and methodological analytics competence

We follow the basic rule that the company should have a strict plan of how the database and the competence of the analytics team are developed together. This cannot be planned exactly, but it has been proven that a systematic approach is important and sufficient. The company must ensure that key results and findings from the analytics projects are not only archived, but that they are accessible to all employees and flow directly into the further development of the analytics strategy and team. This also means that the opinions of employees who are directly or indirectly involved are taken into account. That makes the analytics initiative better quality and more inclusive for all employees, which in turn leads to more participation and greater success.

Very few digital champions had or created a high-quality database at the beginning of their programs before they started analyzing it. The approach of first collecting a large amount of data and then starting an analytics program on this basis causes costs and delays in most cases. Analytics champions who manage their data like a product have resounding success in the development of the database that is fit for purpose. They invest strategically in the further development of the database and align these investments with the planned use of the data. Database and analytics develop step by step and are coordinated with each other. This includes systematic quality assurance, the quantitative measurement of the economic benefit of each analytics project, as well as the deployment of standards and usage regulations that are scalable across the company. The responsibility for the data products thus lies almost automatically with the business, not with IT or the analytics team.

Successful analytics companies work to build their own competence in the field of data analytics. This learning includes all departments that work together with the analytics initiative, as it helps enormously, if the specialist departments also have a basic understanding of how data analytics works. Again quoting the McKinsey study, digital champions are 1.5x more likely to employ analytics professionals and they are 2x more likely to bring different areas of expertise together on or with the analytics team. This makes it obvious how analytics competence does not just mean that employees that work on and with the analytics team master the methods; analytics competence also means that the company successfully establishes structures and routines for employees from different areas to work together.


One of the greatest levers for digitization strategies and in fact often for the entire business model of a company is to build up analytics competence in an in-house analytics team. In this way, the organization builds up a sustainable competitive advantage as it is able to find innovative potential in a variety of ways. Analytics champions use best practices that allow them to achieve rapid progress of their initiative with minimal investment. These best practices focus the program on key goals:

  1. Add analytis to the companies core competences
  2. Systematic development of the database
  3. High program speed and agility
  4. Focus on strengthening the company's core processes

The approach of the analytics champions can not only be adapted by other companies, it also makes it possible to remove most obstacles without taking any risks in advance. Analytics champions break complex use cases in smaller projects. They then implement these projects with agile iterative processes, that enable them to learn quickly and feed their learning back into the strategy. Bill Gates has been quote saying "I'm not interested in data for its own sake, I'm interested in how it can be used to make things better." We can only agree with Bill Gates.

Article image
The best way to a good database: Manage Data as Products
Article image
How to turn data analytics into a competitive advantage for your company
Article image
Best Practices für Data Analytics Projekte