Data Analytics is the new healing promise. What can companies learn from Data Analytics in universities? We show procedures, methods and challenges.
Companies have not only been undergoing change since the digital revolution. Only those who have learned from experiences with customers and business partners and have continuously improved products, services and processes have been able to hold their own in the market. And once a trend is overslept, it becomes difficult to catch up with the competition – just as it is for Nokia in smartphone development.
The difference? With the easier availability of virtually unlimited IT resources, for example via the cloud, Data Analytics has now taken over virtually all application areas. No matter how small the company, there are plenty of providers of analytics applications such as customer analytics, retail analytics and people analytics. The promises of these vendors range from a better understanding of the company’s internal processes and higher profitability to the promise of salvation that we will all only do what we enjoy doing.
In fact, Data Analytics can only contribute to a company’s success if it delivers sustainable added value for its customers. That’s why companies should ask themselves what they can learn from learning analytics at universities. Just as customers produce data when interacting with companies, students produce data when interacting. Learning Analytics refers to the collection, integration, analysis and interpretation of this data with the aim of increasing the learning success of students. The focus is on the added value of the “customer”, i.e. the student – and the added value of the “company” depends directly on the success of the customer. Furthermore, we discuss the approach, the methods used and the challenges in Learning Analytics and transfer these findings to companies.
Achieving customer success
In order to achieve the goal of “customer success”, universities proceed in four steps in Learning Analytics. The first two steps are based on past experience:
What has happened? (Description/Descriptive Analytics) First, universities have to collect and merge data. In concrete terms, many data sources have to be integrated: Result data from the examination systems, data on the learning process from the Learning Management System, personal data from the Student Information System (SIS) and data from external sources. The same also applies to companies: internal and external data about clients must be collected, processed and made accessible for the following analysis steps.
Why did it happen? (Diagnostic Analytics) Once the data has been collected, it is time to examine the data. Learning Analytics data does not have the classic four V of Big Data (Gartners Volume, Velocity, Variety and Low Veracity). Nevertheless, it is worth using methods from the field of artificial intelligence (more on this later), because there are many possible explanatory factors for the learning success of students. As with companies, the more comprehensive the data set and the more flexible the method, the more accurate the findings.
The third and fourth steps deal with the future:
What will happen? (Forecast/Predictive Analytics) Once the connections between the influencing factors and learning success have been established, the prediction can begin. How prone is a particular student to diarrhoea? Or for companies: How satisfied is a certain customer? In a second step, universities and companies can derive these answers from the diagnostic results. But beware: different analysis methods can deliver completely different results for the same “customer”. A robustness analysis is therefore an essential part of the prediction.
How should we act? (Recommendations/Prescriptive Analytics) This last step is the supreme discipline, because it derives recommendations for action from the previous analyses. By means of optimisation methods or simulations, universities can determine how a change in the current procedure affects learning success. And companies: Instead of speculating about the effectiveness of certain actions, they can precisely assess their consequences.
There’s no way around AI
The huge amounts of data available to universities through personalised learning, which form the basis of the four outlined steps, can of course no longer be evaluated manually. In order to draw information from the data about the factors influencing learning success, methods from artificial intelligence (AI), machine learning (ML) and deep learning (DL) are increasingly being used. But what is behind these buzzwords? What do universities need and what do companies actually benefit from? In his comprehensive overview of AI development “The Quest for Artificial Intelligence”, AI and robotics pioneer Nils John Nilsson writes that AI should make machines intelligent and thus allow them to function appropriately and proactively in their environment. In the simplest case, this can happen through a fixed rule: If a student fails the trial test, he or she is presented with the learning content again. Or: If a customer’s contract is about to be extended, he will be sent offers. But which learning contents are particularly helpful for which students, which offers interest which customers?
This is where machine learning comes into play. These are statistical methods with the help of which machines can learn from data. In essence, the aim is to replace the fixed rule with a data set that allows the machine to learn more or less independently. In supervised learning, the algorithm can access a database that provides the relevant categorization promising or useless for each individual constellation. In unsupervised learning, on the other hand, the computer must recognize conspicuous patterns without a given classification.
Algorithms record exact behavior
Deep Learning is a method of ML that is oriented towards the structure of the human brain: Data is not directly related to the result. Instead, in many individual layers different information is extracted from the data, forwarded to the next layer, where it is further processed and only at the very end is it related to the observed result. This is less complicated than it sounds: For students, for example, the first step could be to record how long a student has worked with certain learning materials. In the next steps, a distinction is then made as to whether the interaction took place only once or repeatedly and directly before the exam or with a longer lead time. DL algorithms can capture students and customers in (almost) all facets of their behaviour and thus suggest the optimal path. Even if the use of DL today is still a dream of the future at many universities and companies – the potential is enormous.
In the technical field, the challenges lie primarily in the area of data sources and the resources required for analyses. For universities, a comprehensive LMS is a promising approach that avoids data silos or isolated solutions. Companies, too, have to rely on integrative approaches in order to be able to collect the entire treasure trove of available data. But even the best data set is useless if resources and data analysis capabilities are lacking. The cloud offers computing power and storage space, and an open architecture of the application used and a large user community offer access to solutions and competencies.
Computing power comes from the cloud
Data protection and EU-DSGVO are on everyone’s lips – and of course any analytics strategy must take data protection regulations into account. For example, data storage in the cloud must be legally compliant. But what looks simple in terms of technical implementation is much more complex in detail. The right to informational self-determination gives students as well as customers sovereignty over their own data. But who owns the data created during the analytics process? Students or customers and universities or companies “produce” the findings together. So who can access the data, when and for how long, what happens after graduation, for example, or after the end of a contractual relationship with the data? Beyond this, analytics processes can also lead to ethical problems. It is said that algorithms make decisions only on the basis of objectively verifiable facts. However, current studies show that algorithms adopt the prejudices of their human counterparts. Insurance premiums go up with the wrong first name, and perhaps even students with the wrong address will be disadvantaged when they are admitted to certain courses of study.
Clear commitment to data protection
If data analytics is to have a real impact and lead to data-based decision models, it must be seen as a strategic task. Leadership and culture therefore play a central role in successful implementation. The first step focuses on the development of data literacy: All participants (employees, students and customers) must understand what data is collected and what happens to them. In the second step, processes must be developed that take into account the concerns of the various groups: students do not want every bad intermediate result and every missed exam to be eternally after them. And customers want to be able to decide for themselves how to use their data. This is where the advantages of analytics, full transparency and a clear commitment to data protection in accordance with EU-DSGVO come in: storage only with an opt-in that can be revoked at any time. Employees may fear making themselves superfluous and being restricted in their freedoms. On the contrary, analytics can create room for manoeuvre and help to concentrate on effective measures. In short, data-based decision models cannot be implemented without appropriate internal processes, consideration of the interests of all parties involved and a clear anchoring in the culture.
Conclusion: Data Analytics helps companies to make better decisions and score points in competition. In order to ensure the sustainable success of a company, however, the customer and not the company itself must be at the centre of attention. This is why companies can learn from universities when introducing and implementing Data Analytics: Learning Analytics places the success of students at the centre of strategy, procedure and methods.