You are on page 1of 2

The Reason for The Failure of Forecast and Its Solution

- Focusing on the Coconut Uncertainty


Business Administration

2022123268 Park Seoyeon

1. We can think of uncertainty in two ways: Subway Uncertainty, Coconut Uncertainty.


When a person's commuting time is statistically analyzed, almost all values are expressed
on a normal distribution. There are no extreme values. In other words, it is possible to
explain mostly in sophisticated statistical methods around the average. This normal
distribution graph represents Subway Uncertainty. This effectively models commuting
time, indicating the uncertainty of going to work earlier or later than average.

On the other hand, Coconut uncertainty, unlike Subway uncertainty, is a freak


happening that cannot be explained by a statistical model (normal distribution). Suppose
a person goes to Thailand on vacation and is looking for a cool shadow under a palm
tree. At this point, a coconut fell 'suddenly' over the man's head, and he had a terrible
accident that nearly killed him. These extreme values that are unpredictable and do not
have any regular rules are called coconut uncertainty. Accurate predictions often fail due
to exceptional values such as coconut uncertainty, which deviates from sophisticated
statistical methods.

2. Rather than focusing on planning for accurate predictions, managers should focus
on making emergency plans for various possibilities. No matter how elaborate plan is
made for prediction, it is impossible to predict when, where, and how another coconut
will fall. It is therefore essential that they have emergency plans that can protect
company from a variety of unexpected possibilities.

During making emergency plans, managers should deal with uncertainty (especially
coconut uncertainty) with Three A's in mind. First, managers should ‘accept’ that they’re
operating in an uncertain world. Rather than believing their own predictions, they should
remember that there are many unpredictable things around them. Second, managers
should ‘assess’ the level of uncertainty they face. They should make predictable Subway
uncertainty a statistical model by all means. Accepting the existence of uncertainty makes
it easier to accept information previously ignored as extreme value, which increases the
accuracy of data analysis. Third, managers should ‘augment’ the range of uncertainty. By
doing so, they can cope with more extreme values. Managers should stop believing in
their predictions and develop a plan to be sensitive to coconut uncertainty by referring
to Three A's.

3. In the post-pandemic world, companies are in a completely new business


environment. As customer behavior and expectations change rapidly, companies are
responding to changing demands by introducing optimized supply chain approaches and
cloud-connected business networks. Demand forecasting provides information to key
operational processes such as DDMRP, inbound logistics, manufacturing, financial
planning, risk assessment, and forming supply chains.1

At this time, I propose to use data science technology to demand forecasting. Data
Science technology is a technology that combines mathematics and statistics,
professional programming, advanced analytics, artificial intelligence, and machine
learning with specific subject-specific expertise to identify actionable insights hidden in
an organization's data.2 Using this technology, data can be collected specifically and
systematically. Therefore, it is expected that the data error rate will be reduced compared
to the previous one, which will greatly help reduce Subway Uncertainty. As discussed
earlier in Three A's, it is necessary to create an environment that reduces Subway
Uncertainty as much as possible to effectively respond to Coconut Uncertainty.

For example, looking at the case of a food distributor, we elaborately predicted


shipments (demand) by introducing machine learning from the method of calculating
shipments (demand), compared to the experience and intuitive judgment of the person
in charge. Time series analysis was possible because past shipments according to the
date were written in the company's food data, and based on this, the goal was to predict
appropriate shipments over the next 10 days. As a result, the forecast error rate for 10
days came within 0.9%.3 Therefore, as confirmed in this example, using data science
technology will increase understanding of rapidly changing consumer behavior and the
business environment after COVID-19 and, in conclusion, achieve effective demand
forecasts.

1
최신 공급망을 위한 수요 예측, https://www.sap.com/korea/insights/demand-forecasting.html
2
‘데이터 사이언스란?’, https://www.ibm.com/kr-ko/topics/data-science
3
머신러닝을 활용한 수요예측 사례와 KPI 모델링 방법, https://davincilabs.ai/blog/?
q=YToyOntzOjEyOiJrZXl3b3JkX3R5cGUiO3M6MzoiYWxsIjtzOjQ6InBhZ2UiO2k6MTt9&bmode=view
&idx=11461691&t=board

You might also like