During the pandemic managers have had to make tough decisions, many of them strategically vital to the performance and survival of their organisation, in a particularly unpredictable environment.

These are often complex decisions requiring a calculation of how actions taken today could affect the results of actions taken in the future. With difficult trade-offs to consider, the route to the best long-term outcome can be hard to find. Indeed, the evidence suggests that managers are not good at understanding and evaluating these types of situations. Fortunately, there may be ways for them to improve.

Dynamic decision-making tasks are those where taking an action today changes the pay-off of the same or other actions in the future. For example, when an internet platform adopts a push for users that seems costly in the short-term, but ultimately adds value to the service and firm. Or if someone who hates exercise resolves to go for a run every morning, which eventually leads to them becoming fitter and enjoying their exercise more.

For the manager, this type of decision-making features in many aspects of corporate life, from product development and process improvement to pricing strategies.

These decisions present an interesting learning challenge for managers. Because results in the future are the product of a dependent series of events, it is difficult to detect which actions are key to producing a good outcome. 

It is a learning challenge that managers appear to be failing, given evidence from a wide range of studies covering activities that include supply chain management, resource management, allocation and competitive strategy. Similarly, in experiments designed to create micro-worlds that mimic real-life business situations, participants tend to make high short-term returns but low long-term profits, compared with possible returns, even with the benefit of managerial experience.

To better understand why managers perform badly we ran an online experiment which reduced the dynamic decision-making problem to a more basic form. We used a well-known decision-making task, known as the Harvard Game, that involved a series of choices between two options (in this case pressing a green or blue button 500 times).

Pressing a green or a blue button produced a numerical pay-off, with the size of the pay-off determined by the percentage of blue or green button presses over the previous 10 choices. Participants had to figure out the best way to maximise their pay-off over the duration of the task. By varying the information given to participants we hoped to gain some insight into how managers approach dynamic decision-making tasks, what factors prevent them from arriving at the best strategy, and how their performance could be improved.

Taking into account the information provided to participants, the results seem to exclude several plausible explanations for poor performance. It did not, for example, appear to be due to uncertainty about the basic nature of this type of decision-making (that pay-offs depend on past actions over a particular period), nor a lack of information needed to grasp the structure and dynamics of the task in hand and calculate the optimal solution. Indeed, other than explaining exactly how to maximise performance, there was little more to tell participants.

Yet, even armed with this information, participants only chose the best approach (always blue, except near the end of the task) 68 per cent of the time, some way below the optimal 100 per cent. Furthermore, providing participants with full information and then asking them to think analytically about the problem and then suggest a solution (without being distracted by any button pressing), produced worse long-term results.

How can managers make better strategic decisions?

It seems that, when faced with a complex dynamic problem – even the comparatively simple version in the experiment –the most common decision-making strategy is what can be described as 'muddling through'. Managers try on different actions for size, assessing current and past performance, and select the approach that seems to work best.

This non-systemic thinking approach is far from optimal, especially when there are many possible combinations to try, and unlikely to find the best solution, other than through luck.

An alternative and potentially more promising approach, for example, would be to evaluate actions depending on the values they are likely produce in the short and long-term and assess the trade-offs involved. But this does not seem to be the method commonly adopted.

Even when participants were given help in understanding how their choices affected future outcomes, they still performed relatively poorly. This suggests individuals may need more than a simple explanation of the consequences of their choices to improve their dynamic decision-making.

The findings may not seem that encouraging for managers engaged in dynamic decision-making tasks, but it may still be possible for managers to improve their long-term performance on these tasks, even if only marginally.

A better understanding of the nature of dynamic decision-making tasks and the basic principles involved may prove useful. For example, it is easy to understand why managers fail to continue with a particular option if early results provide a pay-off that is less rewarding than alternatives. However, appreciating the nature of dynamic decision-making tasks may encourage the persistence necessary to obtain a better outcome long term.

There is some evidence that managers who have had the opportunity to learn by exploring a range of different actions at length, rather than being under pressure to produce immediate results, can learn more about the best strategy. This route to improving understanding, which is likely to encompass poor outcomes along the way, is costly. However, organisations and managers could tap into the experience of others who have already been through this process.

Similarly, some research shows that computer-based business simulation exercises involving complex organisational challenges can help to sensitise managers to the structure and dynamics of these types of problems. Although there have also been studies that show how managers can struggle to take lessons from one specific problem-solving situation and apply them effectively to other circumstances.

In addition, there is potentially some benefit in placing a value on possible outcomes that are not easy to gauge, in a similar manner to classic strategy tools such as the balanced scorecard, and embedding them in the metric systems used by the organisation. This may make it easier to evaluate both intermediate choices and long-term pay-offs. It may also help with setting out the structure and dynamics of a particular dynamic task in its simplest, most easily understandable format.

The challenge presented by these dynamic decision-making tasks should not be underestimated. This type of thinking is far from intuitive. That said, adopting some of these suggestions may benefit managers.

Furthermore, it was also clear from the findings that while some luck is involved in achieving good results long term, and although most people struggle, a few managers are just likely to be better at evaluating these types of problem.

Given the importance of these decisions, organisations with managers that understand and perform well on these tasks – something that can be assessed – may well have a competitive advantage over their rivals.

Further reading:

Rahmandad, H., Denrell, J. and Prelec, D. (2021) "What makes dynamic strategic problems difficult? Evidence from an experimental study", Strategic Management Journal, 42, 5, 865-897.

Liu, C., Vlaev, I., Fang, C., Denrell, J. and Chater, N. (2017) "Strategizing with biases : engineering choice contexts for better decisions using the Mindspace approach", California Management Review, 59, 3, 135-161.


Jerker Denrell is a Professor of Behavioural Science and teaches Quantitative Methods for Business on the suite of MSc Business courses.

For more articles on Strategy and Organisational Change sign up to Core Insights here.