Blog: Friday, 11 November 2016
Companies need accurate forecasts to keep stock levels just high enough to supply every customer demand and low enough to keep inventory costs in check. But forecasts are influenced by human judgement – and humans are steered by biases. PhD Researcher Clint Pennings discovered that there are essentially two types of forecasters, each with a specific personal perspective on their own errors of judgment, and whose actions are independent of their role in the company. When collaborating, ‘chasers’ and ‘smoothers’ can cancel out each other’s biases, which leads to more accurate forecasting in general, he also found.
Forecasters in sales, marketing and operations often study past sales figures and compare those to their own forecasts to estimate future demand and to find out how accurate they were, says Pennings. How they translate the gap between their expectations and reality into their next forecast can be explained by the type of forecaster they are. Pennings discovered this after running experiments on two groups of subjects: on 357 students and on 72 forecasting professionals.
In the experiment, participants were presented with sales figures and were asked to forecast sales for the next period. They were assigned the role of either a manager of operations or a sales manager and they were told, over 18 subsequent periods, to produce the best possible forecast for their department or for the entire company. To simulate reality, the test group was forced to collaborate with a computerised colleague from a ‘competing’ department or one with a more neutral forecasting style.
Pennings compared the forecasting errors produced – and people’s reactions to them. His analysis revealed that forecasters are not a homogeneous group, as is often suggested in current models, and that their behaviour cannot be explained simply by their departmental role in the company. Instead, every forecaster is either a ‘chaser’ or a ‘smoother’ at heart, Pennings’ study shows.
Around 75 per cent of forecasters are ‘chasers’; they have a tendency to undervalue their own past forecasts and instead respond aggressively to short-term trends, which they are also quick to distinguish. Their bias carries the risk of ignoring long-term trends in demand, and can lead to an overreaction to sudden market changes.
The other 25 per cent of forecasters can be classified as ‘smoothers’. They tend to overvalue their own forecasts and react less dramatically to sudden market changes. Forecasters with this bias are at risk of reacting too late, or responding insufficiently, when markets are really volatile.
Interestingly, Pennings found that every forecaster sticks to their own personal style, no matter what departmental role, type of colleague, or goal they are assigned. This seems to suggest these biases are not the result of learned behaviour, but are a deeply intuitive and persistent in forecasters, the researcher says.
In Pennings’ experiment, smoothers recorded a bigger forecasting error than chasers. But their misjudgements mostly cancelled out the forecasting error of the chasers, making the pooled forecast more reliable.
Pennings’ results show that forecasting bias is not only caused by the department in which people work, but also stems from their innate tendencies when judging their own errors. Recognising this can be a step towards making the company forecasting team more balanced as a whole, and as a result more effective.
Science Communication and Media Officer
Corporate Communications & PR Manager