Algorithms are becoming more and more preponderant in business management. This is due to two developments: the increasing power of computers and the growing number of data stored in different information systems. Thanks to algorithms, more and more processes are automated or greatly facilitated, which allows production and management costs to be reduced and decisions to be made more quickly and with more relevance. Thanks to algorithms, the company thus gains in performance. Understanding and mastering them is therefore a major challenge in an economic context where the search for competitiveness has never been more intense.
When we talk about algorithms, we spontaneously think of algorithms such as Google’s algorithm or translation algorithms. These algorithms act on the treatment of a well-defined type of problem and their performance depends as much on the speed of calculation as on the quality of the response provided. They feed and evolve thanks to the injection and processing of millions, and even billions, of data, making them increasingly efficient.
However, the situation can be quite different in a company. On the upstream side of the algorithmic problem – the data – the data collection can be erratic and their quality not always constant. The data is not always easily accessible either, because it is scattered in different unconnected systems. On the downstream, i.e. the exploitation of this data, things are also different from systems such as Google’s algorithm. Indeed, the company has to optimize processes that are distributed both in time (time horizons) and in space (services and departments). This leads algorithm developers to design and adapt a large number of different algorithms to fit the needs and realities of business management. Here are a few examples:
– Algorithms that detect errors and anomalies in data collection and, in turn, algorithms that repair or correct these errors and anomalies.
– Algorithms that adjust certain management parameters over time (e.g. a standard task execution time) or according to the evolution of micro or macro-economic contexts (these contexts are themselves represented by models that are fed by data captured inside or outside the company); these will be used, for example, to dynamically adapt sales prices or production quantities.
– Algorithms that enrich data based on given contexts (e.g., to supplement basic information and create more complete information or information that is more easily interpreted by decision-makers or other systems).
– Algorithms that trigger actions at given moments or according to given contexts (e.g. to orchestrate a sequence of actions in an automatic or semi-automatic way, for example to coordinate the schedules of different departments).
– Algorithms that optimize planning results based on management objectives translated into KPIs, while taking into account capacity constraints on resources and project priorities.
– Algorithms that suggest contextual modifications to improve KPIs (e.g., changes to the resource portfolio or product mix or project prioritization)
Algorithms are characterized by their diversity, a diversity that is linked to the countless fields of application they can cover. It is therefore important to understand which algorithms to implement according to the available data, their quality, but also the purposes of their use in terms of management automation and decision-making support. A cost/benefit analysis must obviously frame this study in order to avoid developing or proposing algorithms that would not be appropriate for the field of application concerned.
This is why, as much as specialists in algorithms and computer science will be needed to develop and implement the solutions, they will have to be designed by business consultants experienced in the management and optimization problems that these algorithms will have to solve.
Such guidance is all the more crucial since the companies that will make judicious use of this new class of expert resources (algorithms) will undoubtedly be the leaders of tomorrow.