In recent decades, the significant and rapid development of technologies and information technology has brought business activities in every sector towards an “intelligent” approach driven by data. This has allowed us to increasingly shift the focus to the customer and his needs and requirements. In this context, information becomes a strategic resource to support the decision-making process and is considered an actual corporate asset.
It is in this context that Decision Support Systems (DSS) are placed, supporting software for decision-makers capable of providing helpful information for decision-making processes in a fast and versatile way, thus helping them to have complete control of their business context and make informed and data-driven decisions. It is not just the intuition, experience, or ability of the decision-maker that determines the success of a strategy or the very survival of a company.
Still, every decision must fill a need, and, above all, the speed with which a decision is made is fundamental to reducing the problem-solution gap. The objective of a DSS is to collect, transform, and disseminate information in an “intelligent” way to help the user make decisions suited to their context. DSS systems, in fact, are not created to replace the decision maker but to give him support: the decision is obtained by combining human evaluations with the information processed by the system.
What Is A DSS – Decision Support System
Wanting to give a formal definition, we can say that DSS, Decision Support Systems, are software systems that make a series of data analysis functions available to the user, the decision maker, through the application of mathematical and statistical Machine Learning models in a quick, interactive and straightforward way, with the aim of increasing the efficiency and effectiveness of the decision-making process. Starting from this definition, some aspects of DSS systems emerge that characterize them from the point of view of the service they offer to the decision-maker:
- Ease of use, intuitiveness, and flexibility of the interface;
- Interactivity of the analytical environment and interface;
- Effectiveness and usefulness of the analytical models and data of interest;
From an application point of view, to develop a decision support system, there are basically four main phases:
- The “Intelligent” phase, the one in which analysts, data scientists, and domain experts work together to define the list of valuable data and information coming from inside and outside the company. Furthermore, in this phase, the real problem to be addressed is identified, and the data is, in fact, selected based on it;
- The “Design” phase is the moment in which we proceed with the analytical experiments and end up with the construction of the analytical model (often not just one) from which various possible solutions are generated;
- The “Choice” phase, an evaluation of the different solutions developed is carried out. This phase is aimed at choosing the optimal solution (one or more than one) with respect to the business context and the problem you are trying to address. Field testing activities are included in the choice process;
- The “Implementation” phase, in which the DSS is created based on previous decisions and evaluations by implementing the chosen solution.
Finally, a fifth phase could be named that relates to feedback from end users, the moment in which decision-makers begin to use the DSS, implemented in the company, in their daily lives, not only in a theoretical way but to make “true” decisions. Their evaluation is essential to validate the previous phases and be able to approve the DSS or correct it where necessary.
From the implementation point of view, during the phases above, we find ourselves facing some technical aspects. Even if they may seem unrelated to the ultimate goal, these are choices that must not be overlooked or postponed but which are indeed essential to evaluate immediately in order to avoid finding ourselves having implemented DSS systems with shortcomings that are difficult to resolve after the fact:
- managing large quantities of data: it is a common challenge nowadays in which the diversity of data sources and Big Data raises the question from a technological and architectural point of view of making choices that are adequate for the expected amount of data and also the type of analysis you want to apply;
- access different data sources on different platforms: the problem is closely connected to the first. In fact, we are increasingly faced with situations in which, for various reasons, the data is found on different sources and platforms, and it is essential to have a complete overview of all of them to integrate them best and identify the beneficial information. Critical issues of this type can be resolved by using Data Virtualization tools to create a single data access interface, abstracting from their physical location and access methods;
- guarantee access to multiple users with different permissions because it is true that a DSS system must also be democratic, but this does not mean allowing everyone to see everything. It is necessary to provide a profiling policy and manage access and analysis permissions to the various information. Aspects related to access and, therefore, to the control and management of data can be quickly resolved thanks to Data Governance, the discipline that deals with bringing order to data to allow complete control of it;
- manage historical version of data: analytical models require historical datasets to be able to be trained adequately. On the other hand, these are data that are often not queried directly precisely because they relate to past periods and are no longer attractive. Managing historical data, therefore, has a significant relevance both in terms of analytical feasibility and to address query speed requirements.
How Do You Evaluate The Goodness Of A DSS? Some Guidelines
The creation of a Decision Support System must respond to a series of specific requirements linked to the business context, the characteristics of the decision-making processes, the problems you want to solve, and the user’s needs. It is on all this that the goodness of a DSS system is based, understood as its ability to adapt to the context and support decision-makers in their activity.
For obvious reasons, these requirements cannot be the same for all companies, but we can define guidelines to collect the focuses to keep in mind when evaluating the quality of a DSS system. Flexibility is undoubtedly part of these fundamental requirements: a DSS must be able to adapt to various types of problems, multiple types of decisions, and various types of data, which consequently imply different methods of interaction, processing, and mathematical-statistical analysis, as well as different typologies of users who find themselves using the DSS system.
There are also two main categories of DSS evaluation metrics based on the concepts of effectiveness and efficiency. Effectiveness is understood in its original meaning as the ability to achieve the set objective. In this case it represents the indication of the fact that the appropriate actions are being taken to create accurate information support for decision makers. Typically, the effectiveness of a DSS is expressed as the degree of conformity to the purposes for which it was designed.
DSS systems can be considered as systems oriented towards decision-making effectiveness as their goal is precisely to improve the information content linked to the achievement of corporate objectives and potential business results. Efficiency evaluates the ability to achieve the objective using the minimum necessary resources. It is based on the power of a DSS to “obtain the best possible result with the minimum effort.”
This is a metric that indicates whether the setting chosen for the system is also the optimal path and, therefore, whether the DSS system is designed to optimize the resources available to respond to the company’s needs practically and helpfully.