The knowledge pyramid: How data and information become knowledge

Knowledge staircase – an alternative model

The knowledge staircase is an extension of the knowledge pyramid. It focuses specifically on the needs of companies and the analysis of big data. The individual levels correspond to those of the knowledge pyramid, supplemented by a fifth: skills. However, instead of levels, this model is based on steps. The goal is to systematically create value from data: the company successfully implements the recommendations for action and thus achieves its business objectives.

The knowledge ladder consists of five steps: data, information, knowledge, action, and skill. (Source: Own representation)

Data, information, and knowledge—what these terms mean

Within the knowledge pyramid, data describes measurable quantities that are stored digitally, such as a specific sequence of numbers. On its own, this does not initially provide any information about its meaning. However, when the data is supplemented with additional attributes or metadata, information is created: for example, the sequence of numbers is the item number of a product that customer X has purchased. Linking this information with context and empirical values ultimately creates knowledge.

Stay up to date on personalization: Sign up for the epoq newsletter.Register now!

If the purchased product is, for example, a soccer ball, the item number of the purchased product can provide an indication of the customer's preferences. With this knowledge, you as a shop operator can display personalized product recommendations in the form of matching items (soccer shoes, jersey, ball pump) that are of interest to this particular customer.

How knowledge is created with the help of artificial intelligence – examples

There are two types of artificial intelligence (AI): knowledge-based AI or expert systems, and data-based AI.

Knowledge-based artificial intelligence

Knowledge-based AI, also known as expert systems, supports human experts in their decision-making. To do this, the experts first capture the necessary knowledge and make it understandable for the computer. Based on this, the system develops solutions to problems, which are then presented to the human user in an understandable form. These expert systems are less intelligent than data-based AI.

Instead, decision-making here takes place via fixed "if... then..." rules, which the programmer uses to communicate their knowledge base to the machine. This makes the machine less flexible and independent in its decision-making. One disadvantage of expert systems is therefore that the knowledge transmitted can become outdated.

Example of an expert system: With top seller recommendations, you as the shop operator specify exactly which products you want and tell the system clearly, "Give me the products with the most purchases in the last x days." As an expert, you believe that frequently purchased products are a good choice for your recommendations.

Top seller recommendations are an example of the use of an expert system.
(Source: Screenshot from gepps.de)

Data-based artificial intelligence

In this case, the AI must be trained using algorithms and statistical models in order to draw relevant conclusions from a set of data. There are various approaches to this training process, such as supervised learning, unsupervised learning, and reinforcement learning. Once the AI has recognized the patterns and dependencies between the data, it can make decisions independently. However, the AI's decision-making process is difficult to understand. It is also heavily dependent on data quality and is not suitable for every task. For example, aesthetic features, which play a major role in product recommendations, are difficult to convey to AI.

Example of data-based AI: If you, as a shop operator, want "good alternative recommendations," you can submit a request to the AI and tell it what you want. The artificial intelligence decides which recommendations to make based on the data. So you just give the AI a little help by specifying which strategy (in our example, "alternatives") should be used. This also involves a little bit of expert knowledge.

Product recommendations in the form of alternatives are an example of the use of data-based AI.
(Source: Screenshot from fackelmann.de)

Knowledge pyramid and AI – the basis for personalization

As the basis for AI engines, the knowledge pyramid also plays a crucial role in personalization in e-commerce. As already explained, the process of making data-driven decisions is divided into three parts. In the following, we will apply this process once again to AI-supported personalization in e-commerce.

Stay up to date on personalization: Sign up for the epoq newsletter.Register now!

When we talk about big data in e-commerce, we mean all data collected in the context of online shopping. This includes click and purchase behavior, which can be enriched with data from the shop system or merchandise management system.

Information is now the data that is crucial for selecting personalized customer recommendations. The art of recognizing patterns and correlations in data in order to extract useful information for the online store is called data mining. Its use results in smart data, which has a significantly higher information density than the previous big data mass.

Knowledge is now the ability to make decisions based on information and experience. This experience can be gathered not only by humans, but also by machines. For example, reinforcement learning can be used to make decisions about personalization based on given information.