Imagine that we are able to predict the future. No, I do not mean predictions on the level of Apocalypse, another end of the world, or Nostradamus. I mean a simpler future that is, however, more probable as something based on science. For instance, it may refer to the development of our business. An increase in demand for a product or product groups may change the operating method of the marketing department and sales methods with regard to a specific target group. If we ran a shop selling ventilators, we could make use of average annual temperature ranges to determine when it is profitable to order larger quantities of goods. Isn’t that smart? And what if we talk about global business?

The way in which analytics developed around the world is quite simple. At the beginning, data were collected and operational reporting was created on their basis. Another step was analytics answering the question what happened (in the past) that we received a specific result. This was followed by multidimensional analyses in the OLAP data structure. These complicated spreadsheets made it possible to process rapidly growing quantities of data within the scope of their mutual relationships. Today, instead of asking what happened, we ask why it happened. We are entering the era of predictive analytics.

Looking forward

In further periods, we may deal with prescriptive analysis, which will help to answer the question what a business should do. Equipped with its own artificial intelligence, it will not only deliver predictions but also prompt the best solutions for the existing situation. For example, it will assess what kind of customer will buy a product in our shop and what we can offer as a supplement of the offer to that specific user. We will also learn how products are connected and what relationship exists between goods and customers.

It is already beginning to happen

For the time being, artificial intelligence does not do this for us, but we can obtain a great deal of information from data we have collected. If Nokia had asked itself a question and analysed the available data at the proper time, today it might still be the leader in the production and sale of mobile phones. The beginning of the end for the Finnish giant dates back to 2007, when Apple presented its first iPhone with a touch screen. It is worth emphasising that the first global patents of this solution – and also the first finished products – were elaborated by the Finnish. Another nail in the coffin was the fact that Scandinavians did not develop their Symbian operating system, which made users feel uncomfortable with the equipment of this company. However, Nokia made more mistakes in the prediction of trends: from ignoring the fashion for phones with a flap to the lack of an offer for younger generations of users. Nokia’s overall anachronistic approach to business resulted in stock exchange decreases and the sale of the company to Redmond-based giant Microsoft.

We should not be like Yahoo

The history of Yahoo and AltaVista search engines is a similar example. While Yahoo may still sound familiar to us now and then, AltaVista was rather doomed to oblivion (it was closed in 2013). But let us go back to 1998. Two young students of Stanford University make a proposal to AltaVista and Yahoo on the sale of their search engines to Google, which contains a modern page segregation algorithm PageRank. The asking price is only 1 million USD. Both companies refuse. In 2002, Yahoo realises that its decision was wrong and evaluates Google at 3 billion USD. However, Google sets the price at 5 billion USD and Yahoo refuses to buy. In 2008, the offer of purchase of Yahoo is made by Microsoft, which proposes the amount of 40 billion USD. Yahoo refuses. In 2016, Yahoo is sold to Verizon for merely 4.6 billion USD. Now let us imagine that we have data from which we can learn about trends and the level of use of our solution. Why do we fail to make use of them and repeat the case of Yahoo?

Predicting the future

On the example of the mobile phone operator, we can understand how modern predictive analytics and optimisation works on the basis of SAP Analytics Cloud. Imagine that we want to contact customers who will buy our product or service. How to choose the customers who are interested in our offers from the database?

– If we have information who chose our service or product in the past, we may assume with certain likelihood who will be interested in that service or product again. Therefore, we must divide customers into two groups: those who have bought something and those who have not bought anything yet. Then, on the basis of this valuable information, we can create the classification rule and build a predictive model. For that purpose, we need to deliver more information: demographic, behavioural and any other data that may affect and improve the quality of prediction. While testing our model based on test data, we make sure it picks the relevant customers. When we are already convinced that the results are right, we can apply this model to all of our potential customers. And, thanks to such information, we send telemarketers data concerning subscribers who should be most easily persuaded into buying from us, tells Michał Bekus, SAP Analysis Expert at Hicron.

3 directions of prediction

Directions connected with prediction needs are diversified. On the one hand, it is an important direction for the development of small companies and startups; on the other hand, even stock exchanges should be interested in the development of this analytic method. In this context, we can distinguish 3 target groups of analytics:

  1. Data Scientist – analysts of Big Data. This creates the possibility of using analytic predictive methods created by SAP on the premises and, at the same time, for the use of statistical knowledge. Analysts also use automatic prediction libraries to accelerate the research process. For example, they can use the SAP Analytics Cloud in the first place to check that the predictive model used there shows interesting relationships and then build their own predictive model of higher quality.
  2. Business users – depending on the experience and skills of analytic users, they may use automatic prediction models and visualise data delivered from various systems. On their basis, users may draw conclusions and make more adequate business decisions.
  3. R language programmers – prediction embedded in business applications that helps to expand the scope of analytics and to create advanced visualisations useful in an analysis.

What does SAP Analytics Cloud give with regard to prediction?

Let us imagine that we have a sales company distributing clothes and accessories. The management board of the company would like to increase sales results and, for this purpose, needs to elaborate a new market strategy. The analysis department receives a request to present a report answering the question which factors have an impact on revenue on the basis of sales data. The report must be based on a certain type of data. We can deliver them in three ways:

  1. From flat Excel files.
  2. From a source system (e.g., S4HANA or BW4HANA) or various other databases.
  3. From the data model that we created earlier.

From data collected by the system, we can learn which showrooms achieve good or poor results and understand the reason for this – e.g., customers’ opinions, the geographic location, the size of the department or the ease of purchase. The machine learning technology embedded in the analytic program allows us to match selected information to the analysis and generate a visualisation on their basis to reach conclusions more easily. The system is smart enough to notice sales anomalies and exclude them from the report. Thus, the quoted ventilator seller will know that the summer of 2018 was excluded from the analysis, because this was a period of unusually hot weather in Poland. Having such information, smaller and bigger entrepreneurs have a chance to stay ahead of the competition and follow current trends.