22.11.2019

Take the first step to become a data driven company

How to find use cases for modern technologies such as machine learning.

Comprehensive digitization and networking mean that huge volumes of data are generated in companies. And they want to use this data with the help of intelligent processes and machine learning to generate added value. But many projects fail because companies neglect the strategic relevance of the entire use case or fail to consider it at all. This is where a comprehensive data strategy can help. It enables companies to successfully shape their digital future, make better decisions and optimize processes.

New business projects often start with the words “I’ve got an idea: Why don’t we ...”. An employee or the CEO has developed an idea for a use case or was inspired to do so—be it at a conference, through an article or a conversation with a customer.

In fact, many projects start from such spontaneous ideas. And it is generally also not a bad starting point for gaining first experiences and learning lessons from them. But many ignore the fact that they may have limited or sometimes not the right resources in terms of people, tools, data and time available.

Possible consequences: The implementation fails, comes to nothing or turns into a never-ending story. Once the use case is implemented, it offers little added value and is ineffective. Therefore, it is not put into practice. Accordingly, companies should clarify in advance whether the necessary resources are available and the use case will increase efficiency and be strategically relevant. A few successful examples from various fields of application show how this is done.

Implemented use cases

Football club sells more tickets

Football club sells more tickets

Ranking lower than before, 40 percent of the club’s fans don’t extend their season ticket. Together with data specialists, the club developed a scoring model to identify, classify and prioritize suitable leads and assess the risk associated with season ticket holders. By using the identified patterns and addressing their fans in a targeted manner, the loyalty of season ticket holders could be increased and new season tickets sold.

Pharmaceutical company reduces rejects

Pharmaceutical company reduces rejects

A pharmaceutical company had an unusually high rate of rejects in a specific month. The machine-learning-based assignment of products to defect categories led to the early detection and avoidance of faulty rejects as well as 75 percent cost savings through an improved quality assurance process.

Semiconductor manufacturer identifies pseudo-errors

Semiconductor manufacturer identifies pseudo-errors

15 percent of DUTs of a semiconductor manufacturer are classified as faulty although they function correctly. Machine learning helped to avoid 90 percent of the previously detected pseudo errors. This led to annual cost savings of 250,000 euros, because the manufacturer no longer has to carry out time-consuming follow-up inspections and discards fewer products.

Automotive supplier taps into new markets

Automotive supplier taps into new markets

An automotive supplier wanted to know how the company could make meaningful use of data. In a two-day workshop, a use case was developed to unlock sales potential in new markets, including the required resources such as data sources, skills, tools and partner companies. The implementation took place in the context of a three-month project, at the end of which the exact sales potentials at product level in three markets were identified. On this data-driven basis, the management ultimately decided to introduce carefully selected products with high sales potential. 

Recommended approach

These examples show that a company should first develop suitable use cases based on existing business models, strategies and processes and evaluate them in terms of added value before implementing them. However, the starting situation varies greatly from company to company. This is why use cases need to be customized and developed for a specific purpose. They usually form the starting point for deriving a holistic long-term data strategy, including data governance and data management. However, excessive expectations also need to be addressed in this context. The management, for example, wants to immediately achieve optimization through AI, but the data quality, IT structures and the skills of the employees are not sufficient for this. To minimize risks, we recommend a step-by-step approach based on a specific use case and using the Design Thinking method.

Step 1: Generation of ideas (diverge)

Step 1: Generation of ideas (diverge)

This is about understanding the business model, the corporate strategy and the level of analytical maturity. Involvement of participants from different departments and decision-making levels to highlight the goals to be achieved and identify existing challenges.

Step 2: Prioritization of ideas (converge)

Step 2: Prioritization of ideas (converge)

Specific use cases are identified based on the ideas. These cases must then be prioritized according to added value and complexity. The use cases can contribute to the current business model or develop potential new ideas for business.

Step 3: Specification and development of the different use cases

Step 3: Specification and development of the different use cases

Which resources (data, skills, tools/technologies, processes) are required to achieve the desired added value? The data and skills landscape per use case needs to be analyzed in this context to identify possible deficits and remedy them accordingly.

This approach does not require wizardry and can be implemented through a well-managed two-day workshop. The workshop will identify data-driven approaches for each company and, on this basis, develop a holistic data strategy.

Practical tips

  • A company should start with their goals or challenges and define the added value a use case should deliver. Only then should they assess which data and other resources and processes are required for implementation. Not the other way around!
  • The right people should be involved right from the start: They must adequately represent the depth and breadth of the company.
  • Expectations need to be addressed correctly! Exaggerated expectations should be put into perspective early on, while underestimated opportunities should be supported.
  • Overall, a reflective facilitation process that is adequate for all stakeholders needs to be ensured.
     

Key questions before kick-off

  • Which use cases is the company is busy with?
  • What target group (internal, external/sales, marketing, production/decision makers, experts) do the use cases refer to?
  • What added value do they provide with regard to this target group?
  • How do these use cases contribute to the strategic goals of the company?
  • How are successfully implemented analytics projects measured?
  • Which projects have failed?
  • What is the level of maturity? What skills, data, tools, technical expertise does the company have?