Learning from the past in order to master the future – for companies this means: Using Predictive Analytics to estimate the probability of future events in order to draw relevant conclusions for the future.
Text: Felix Raymann, Images: Alamy, Keystone, Strandperle, 20
Getting a crystal ball or relying on the services of a fortune teller is not really an option for most companies looking to improve their HR performance. So how can you find out which valuable employees are about to leave their jobs? How do you track down those employees in the company who will reveal themselves to be talents worth nurturing in the future? Predictive Analytics can help achieve these goals: Company data is used to calculate the probabilities of events occurring. The method enables statistical patterns of behaviour to be used to predict complex economic interrelationships.
The lack of skilled workers in certain industries is worrying. The best employees leave before the new generation in the company is ready to take their place. To find a solution to this disadvantageous HR situation, the company has to know how to hold on to its valuable employees and how hidden talents within the company’s ranks can be unearthed.
Not an easy task, because resignations often seem to appear out of the blue. And it is not always the loudest employees, those who are good at selling themselves, whose promotion will add to the value of the company. A company’s employees are its most valuable resource. So it is worthwhile taking a closer look and making use of the available information by examining the relationships between various HR data.
“The basis is made up of the widest range of indicators, such as salary, salary development in recent years, years of service, employee review evaluations and age,” explains Martin Gutmann, Head of Analytics & Data Consulting at Swisscom. For example, the SAP Predictive Analytics analysis software investigates along its “Job termination” row how all of the characteristics in the columns match up. “The more of these characteristics one uses, the greater the chance of being able to detect significant matches and thus make predictions,” says Gutmann. “This is also a way of finding those indicators that are not obvious at first glance, but that still influence a job termination. For example, the evaluation of the employee’s superior and even the evaluation of this superior’s boss,” says Gutmann.
In order to be able to generate meaningful data, the information and raw data must be aggregated. “The cleaning, enriching and preparation of the data accounts for around 80-90 % of the work,” says Martin Gutmann, illustrating this with a simple example: “If you want to make predictions for sandwich sales figures in front of Bern train station, the sales figures from the past give you only very limited information for future sales. However, when you enrich this data with other indicators such as the weather conditions, the day of the week or events taking place nearby at the same time, useful predictions can be made.” The system uses algorithms to find out which combination of indicators is most probable for a specific event to occur.
Knowing in advance how customers are going to behave, when the ideal time is for maintenance, or when fraud is going to be perpetrated, gives a significant competitive advantage and protects against unpleasant surprises. The uses of Predictive Analytics are wide-ranging and encompass all industries. A few examples: In marketing it can be used to determine the target group and the right communication channel. In industry – and especially in machine engineering and IoT projects – accumulated data volumes can be analysed. The focus here is on detecting anomalies in order to make maintenance more straightforward. The process is also applied to invoicing in order to predict customers’ payment practices. If you know that a customer will pay a bill late, but in all probability is going to pay it, you do not send a reminder and the customer experience remains positive as a result.
In the finance sector, Predictive Analytics enables you to optimise your cash flow and make better predictions about it. Banks can improve their credit risk management, optimise their client portfolio or foresee money laundering activities more reliably. “The better a bank understands its customers, the better the consulting it can offer. Instead of making standard offers to customers, a targeted evaluation of the customer data enables you to prepare for specific, relevant offers,” says Gutmann. For example, a large bank in Zurich would conduct an appropriate analysis to find out which customers are very likely not to renew their mortgage. Churn Prediction is used to find the customers most likely to terminate their contracts and then approach them individually. “Along with the characteristics age, gender, type of property etc., other indicators such as ownership of securities, bank balance fluctuations or the use of e-banking are also taken into account,” Gutmann explains.
Evaluating historical data is nothing new as such, and has been used for a long time already in the insurance and retail industries in particular. However, what has changed fundamentally is the data volume and especially the ease of access to powerful algorithms with which predictions can be made. In contrast to Business Intelligence (BI), Predictive Analytics does not merely transform unstructured data into usable information, but also describes meaningful correlations that shed light on future events.
To use Predictive Analytics, however, certain prerequisites must be fulfilled: In particular, there must be data of sufficient quantity and quality. And data protection is an important factor here. It is not merely that the data protection laws must be observed at all times, but also that questions relating to ethics and security need to be asked.
Predictive models are now able to automatically supply themselves with new data at regular intervals, and to incorporate previous results autonomously. This enables a prediction to be continuously reassessed and refined. The recommendation from the algorithms adjusted in this way is analysed and discussed with business experts and then implemented step by step. The tendency is toward end-to-end automation, from the data collection and processing to the preparation and selection of the best models and their automatic implementation in IT systems based on specified criteria.
Risks and trends can be detected early using data analysis. HR data in particular can be used to create detailed reports that enable companies to be proactive in, for example, nurturing valuable employees and also holding on to them. Specifically, Swisscom offers an Ideation Workshop in which companies can work on their own HR application cases.
Would you like to regularly receive interesting articles and whitepapers on current ICT topics?
More on the topic