Planning in the Aerospace Industry: Biases and Insights from Behavioral Economics

The effectiveness of planning has become a critical requisite for the success of industrial companies. In fact, production planning often impacts customer satisfaction and the bottom-line. Under-planning may lead to missed deliveries, while over-planning is a recipe for an oversized supply chain.

That is why it is important to assess the effectiveness of planning and strive to improve the underlying process. In this study, we took cues from behavioral economics to measure irrationality in decision making.

In order to quantify the accuracy and precision of the planning, we took the industrial planning of 4 European aerospace companies and compared it to actual aircraft deliveries over a span of up to 17 years.

The findings give us a first benchmark. Overall, we found low levels of accuracy with a clear tendency to plan above the actual needs—pointing to over-optimism. In addition, when faced with hard decisions, another cognitive bias sets in; we will refer to it as drag-bias.

We also found low levels of precision, suggesting that the trends—changes in production rates—of planning and deliveries are not always aligned. The data shows signs of two other cognitive biases during periods of change: representativeness and ramp-down aversion.

Accuracy #

To determine the accuracy of the planning, we measured the delta (Δ) between planned and actual industrial deliveries in number of units.¹

In this formula, Δ=0% means a perfectly accurate planning—i.e. we produce the exact number of planned units. The following table shows the overall results.

Table 1: Differences in accuracy among companies

The magnitude of the above figures suggests a sizeable potential to improve planning accuracy, at least for companies 1, 2, and 4.

We then measured the evolution of accuracy over time and found that it decreases significantly with the planning horizon. This is shown in the chart below.²

Graph 1: Accuracy degrades with the horizon

On average, the measured delta seems high. In addition, it increases much further during difficult times—or key historical moments in which the aerospace industry saw dramatic reversals of market demand: the financial crisis of 1992-1993, September 11, 2001, and the economic crisis of 2007-2008.³

We wanted to gain an objective understanding of these figures. So we compared the measured accuracy of the planning, in each of the 4 cases, with that of a simplistic planning model. In the model, we copy last year’s deliveries and consider them as the planned deliveries for the current year and the years to come.⁴

We found that the accuracy of the simplistic model is much better than the real planning for companies 1 and 4 (by a factor of 3). The model is equally accurate for company 2, and less accurate for company 3 (by a factor of 2).

This result confirms that—for companies 1, 2 and 4—the accuracy of the planning may be substantially improved.

Over-Optimism #

We found that the delta is positive throughout the studied companies, products, and planning horizons. Consistently, more units are planned than actually produced.

This finding points to over-optimism; a phenomenon similar to the planning fallacy that is discussed in the literature. Whereas planning fallacy refers to the tendency to underestimate how long one will need to complete a given task, over-optimism is the tendency to plan more units than we end up delivering.

Drag-Bias #

Continuing our dive into accuracy, we found a second cognitive bias, which we will refer to as drag-bias;⁵ a similar phenomenon that is discussed in the literature is anchoring.⁶

Drag-bias leads the planning to be more influenced by past deliveries than it needs to be. This behavior is caused by the excessive reliance on past information as a reference to plan future deliveries.

Using the following formula, we examined whether our 4 companies are overly influenced by past delivery figures.

α = [(Planned deliveries) – (Deliveries of year n-1)];

β = [(Planned deliveries) – (Actual deliveries)];

If α < β, then there is drag-bias

And drag-bias, σ = β – α

The results show no evidence of systematic drag-bias. However, this effect appears clearly during periods of crisis and dramatic change in demand.

Every time our 4 companies were faced with hard decisions due to a changing environment, drag-bias seems to have affected their judgment.

Table 2: Drag-bias increases under difficulties

And since, during those same periods, accuracy degrades significantly, drag-bias may be seen as a strong contributing factor to this degradation. Company 3, which achieves the best overall accuracy, seems to pay the price of the sturdiness of its planning process with the highest punctual increases in σ during hard times.

When faced with adversity, we seem to cling to our past. In those moments, our planning becomes closer to last year’s deliveries than to the actual deliveries it is meant to decide. And this results in drops in accuracy.

Precision #

To analyze the precision of the industrial planning, we looked at the correlation between the planned deliveries for the four upcoming years and the actual corresponding deliveries.

Table 3: Differences in precision among companies

A low correlation may cause mistaken decisions regarding the sizing of the company’s internal and external supply chain. For example, low precision, on average, could lead to high inventory levels during production ramp-down, or to undersized suppliers during ramp-up.

However, in our view, a 4-year planning with a correlation of 1 is not the goal either. Within the timeframe analyzed (up to 17 years), there have been periods during which the aerospace industry has suffered from unexpected pressures. In those years, precision degraded abruptly.

A correlation of 1 would have meant an industrial delivery rate misaligned with the evolving demand. Company 3, with an average correlation of 0.51 seems to represent the benchmark within the companies that we have studied—it achieves the overall best precision at the price of steep punctual drops.

Representativeness #

In relation to precision, we analyzed the existence of another cognitive bias—representativeness. Representativeness occurs when the planning extrapolates past trends into the future more than it needs to.⁸ To verify the existence of representativeness within the studied data, we used the following formula.

α: Slope of actual deliveries;

β: Slope of the planning;

δ: Slope of past deliveries;

If |β – α| > |β – δ|, then there is representativeness

And representativeness, ρ = |β – α| – |β – δ|

There is representativeness when we plan a slope that is closer to the previous 3 years than to the actual future deliveries. This formula can reveal irrationality if, by extrapolating past data, the planning is worsened instead of being improved.

In line with the findings from drag-bias, the studied companies seem to be excessively influenced by this bias during moments of large changes in the industry.

Table 4: Representativeness increases under difficulties

An increase in representativeness during difficult years may be a contributing factor to low precision. When in doubt, we reproduce past trends.

Ramp-Down Aversion #

Finally, we identified a fourth cognitive bias from the collected data. We refer to it as ramp-down aversion—an inspiration from the concept of loss aversion which is discussed in the literature.

This phenomenon, affecting decision making, explains how the human mind is biased to weigh more a prospect of losses than the prospect of equivalent gains. Ultimately, it leads us to take larger risks to avoid a loss than to secure a gain.

To measure this bias, we calculated the slope of actual deliveries and that of planned deliveries. We then compared them during two periods: years of increased and decreased delivery rates.

α = Slope of actual deliveries – Slope of planned deliveries; in periods of ramp-down

β = Slope of actual deliveries – Slope of planned deliveries; in periods of ramp-up

Considering an Aversion Index λ = α + β,

If λ > 0, then there is ramp-up aversion

If λ < 0, then there is ramp-down aversion

The following table shows the results.

Table 5: The data shows ramp-down aversion

As we can see from the table above, we tend to underestimate both ramp-up and ramp-down. However, when looking at the values of λ, we can conclude that we are more averse to ramp-down than to ramp-up.

As expected from the previous findings, Company 3 remains the benchmark. Nevertheless, in all four companies, when faced with a dark prospect, we tend to take disproportionate risks.

Conclusion #

This study may serve as a benchmark of the accuracy and precision of industrial planning in the aerospace industry. Although other factors may affect the performance of the planning, our data suggests a substantial influence from four cognitive biases:

• Over-optimism,
• Drag-bias,
• Representativeness, and
• Ramp-down aversion.

At an individual level, these biases are largely inescapable; they are basically innate. However, at an organizational level, we believe that business processes can be adapted in order to lessen their effects.

Companies rely on planning for most business decisions. But we rarely take the time to look at past plannings and the corresponding execution to analyze their quality.

Even when we acknowledge that we live in a complex world, in which irrationality and subjectivity are commonplace, can we afford to leave the effects of such irrationality in our cognitive blind spot?

Surely not. Especially when the stakes are so high and the effort to do it so manageable.

As such, we are strong supporters of feedback loops: companies ought to run systematic studies of past planning performance in order to continuously adjust the underlying processes—thereby improving their output and the associated business decisions.

Authors: E. Dib, C. Criado-Perez #

References #

• Abdellaoui, M., Bleichrodt, H., & Paraschiv, C. (2007). Loss Aversion under Prospect Theory: A Parameter-Free Measurement. ESTP.
• Ariely, D. (2010). Predictably Irrational. HarperCollins.
• Hogarth, R., & Makridakis, S. (1981). Forecasting and Planning: An Evaluation. Informs.
• Kahneman, D. (2012). Thinking, Fast and Slow. Penguin.