Decisions are the driving force behind any organization. Were the decision-making to stop overnight, any company would collapse in no time at all. The reason being that reality is too complex to be adequately planned in advance. The dream of an organization that would continuously churn out a predefined output based on a predefined input, like some kind of machine, has been outdated for quite some time now. Nevertheless, such fictional ideas about how to run things are currently gaining renewed impetus. The hope is that the issue up to now has merely been the lack of necessary availability of data to feed the machine with sufficient complexity.
There are two different modes to data-driven decision-making in organizations. On the one hand, decision-makers act in standardized situations, in which routines are closely linked to the use of or reaction to data. On the other hand, there are non-standardized situations, in which data is also intended to serve as a decision-making basis for goal programs – but without it being clear how this should look exactly.
Routines were already the focus of the discourse around automation during the pre-digital era. Here the case for the well-chosen implementation of ever-improving, more readily available and increasingly abundant data is clear. After all, routines are based on the if-then principle: if, for example, the minimum stock of parts produced in a manufacturing group falls below a predefined level, then the order to re-produce the corresponding parts is automatically sent out. But routines are not just a phenomenon of manufacturing. Even for those who wear suits as opposed to overalls, they are the order of the day. If the contract volume exceeds 500,000 euros, then it must be personally approved by the head of department. Whether it’s ensuring there are enough screws in stock or processing large orders, the logic of routines in standardized decision-making situations is always the same. Predefined information triggers a predefined reaction. The increased availability of ever-improving data in the course of digitalization makes it possible to trigger routines more quickly or to define them more precisely. The more narrowly defined the trigger and reaction are, the less (formal) room for manoeuvre there is for the individual. The logical result of this interlocking is automation, as it is currently being advanced in increasingly complex contexts in the form of algorithms. It would, however, be wrong to think that algorithms already in commercial use today would actively make decisions on a larger scale, i.e. independently make use of room for manoeuvre. As a rule, these are increasingly closely interwoven routines, where the input and output have already been determined in advance.
Current discourse on the role of data in organizational decision-making processes focuses not only on these automation options, but also on non-standardized decision-making situations and the goal programs, which are associated with them. First and foremost, goal programs are characterized by the fact that it determines which goal is to be achieved, but largely leave open how this goal is to be achieved. The choice of means is therefore up to the respective decision-makers, usually limited only by the fact that some potentially conceivable means are excluded by internal and external rules. Data is about to establish itself as the new means of choice for goal programs. In case of doubt, it seems better to base one’s arguments on statistics rather than on one’s gut feeling. Therefore, managers who tend to shy away from taking decisions now see their chance: if complex issues that cannot be squeezed into routines become subject to calculation rather than decision-making, the need for taking the lead in situations of uncertainty will simply disappear – this at least seems to be a widespread hope. Alternatively, data lakes will be created to facilitate serious decision-making on the strategic orientation of the organization. Data, it seems, is the solution for the practice of goal programs, which has always been subject to great uncertainty. However, this perspective overlooks the fact that there are two essential requirements for the use of data in organizations that make this unlikely to be an objective solution to complex problems.
On the one hand, organizations themselves still decide on the selection of data. No matter how vast the data pool, a first assessment of the relevance of some of the data and not the rest, which is also potentially available, is always already programmed into the selection process, which is to serve as a basis for decision-making. Data does not therefore paint an objective picture of customers, markets or one’s own organization; it is always run through the specific filter of an organization thus expressing certain assumptions of relevance and realities. On the other hand, data cannot automatically be used to derive clear recommendations for action. Of course, a 40% drop in sales may indicate a need for action for all those involved. What the appropriate reaction to this environmental monitoring would be, however, is likely to be a bone of contention. Even data must always be interpreted first. Furthermore, the information content of mass data is initially fundamentally low. The information only emerges from the questions that organizations pose – questions whose selection cannot be calculated, but which require a decision by the players involved. This is also true for the issue that data always paints a picture of the past. If and how it is suitable for predicting future developments remains a matter of estimation and is therefore subject to a decision.
Whichever way you look at it, it is clear that data naturally intervenes in organizational decision-making processes, be it closely coupled, in standardized situations or just loosely tied to non-standardized decision contexts. In both cases, however, decision-makers would be well advised to lower their expectations regarding an increasing number of “right” decisions. Whether pre-selecting conditioned reactions to predefined inputs or using data as one of several means for goal programs: data-driven does not mean data-given – decisions can be made without data, but using data always means having to make decisions.
is Senior Consultant at Metaplan Germany.
is Partner at Metaplan Germany.
Stay tuned – we will be publishing the following propositions on ‘when data hits the organization’ in the coming weeks.
3. Informal loopholes don’t disappear when data comes into play – they are covered up more thoroughly!
4. Power games will not vanish due to increased transparency through data – they will shift!