Major domestic demand forecasting failures have shown that there are serious risks to strategic decision-making and public infrastructure investment.
A series of recent failures have exposed the fundamental flaws in current forecasting systems, and have especially highlighted how vulnerable existing demand forecasting methods are in unpredictable situations such as the COVID-19 pandemic.
In the public sector, the failure to forecast demand for large-scale transportation infrastructure projects such as the Singang Line, Yongin Light Rail, and Uijeongbu Light Rail has resulted in a huge waste of public funds.
The actual number of passengers for these businesses was only 30% of the forecast, resulting in the inefficient use of hundreds of billions of won of taxpayers' money. In the private sector, there were a number of management crises caused by failure to predict demand, such as the semiconductor supply and demand failure in the automobile industry and the disruption of manpower management in the movie theatre industry.
The cost of such failure to predict demand goes beyond the issue of mere numbers. In the public sector, inefficient allocation of national finances and over-supply of social infrastructure, and in the private sector, weakening of corporate competitiveness and threats to survival.
Even more worrisome is that these failures are not one-off events but stem from structural problems.
This analysis closely examines major demand forecasting failures that have occurred recently and analyzes the root causes from institutional, technological, and organizational perspectives.
Furthermore, we are seeking innovative solutions, such as the introduction of AI-based forecasting systems, and we aim to suggest overall directions for changes in organisations and systems. Through this, we hope that decision-makers will gain practical insights for the innovation of demand forecasting systems.
Case Study on Demand Forecasting Failures
Over-forecasting of Demand for Transportation Infrastructure and Waste of Funds
Failure to predict demand for transportation infrastructure projects is causing a huge waste of public funds. The second section of the Sinbundang Line had an average daily ridership of 48,000, compared to an expected daily ridership of 166,000.
This is 30% of the forecast, and it clearly shows the failure of a project that has used 780 billion won of public funds.
The case of Yongin Light Rail is even more serious. The number of daily passengers was expected to be 160,000, but it actually only reached 8,000, and the deficit reached 53 billion won in 2014 alone.
Uijeongbu Light Rail and Busan-Gimhae Light Rail are also experiencing similar problems. The fact that the actual number of passengers is less than half of the forecasted number suggests a fundamental problem with demand forecasting.
The background to this overestimation of demand is political influence. Regional development logic and election pledges are distorting objective analysis, which in turn is leading to inefficient allocation of national finances.
Chain reaction of operational risks caused by failure to forecast demand
The semiconductor supply and demand failure in the automotive industry shows how demand forecasting errors can paralyze the entire supply chain. The decision to reduce semiconductor orders has caused the following chain of risks.
In other words, there was a direct loss due to the interruption of the production line. Toyota, Fiat Chrysler, Ford, and Honda had to suspend the production of finished vehicles, which led to a loss of sales.
Even more serious is that semiconductor suppliers have prioritised other industries over the automotive industry and switched production lines to other industries. This has caused structural problems that make it difficult to secure a stable supply of parts in the long term.
Failure to forecast demand worsens a company's cost structure. High-cost emergency procurement was required for sudden parts supply, which was eventually pointed out as the cause of rising costs.
In addition, the failure of large companies to forecast demand also has a serious impact on suppliers. In fact, the suspension of production by automakers has led to financial difficulties for numerous parts suppliers.
Failure to forecast demand in the film industry has led to a shortage of manpower
The case of the film industry shows the devastating impact of failure to predict demand on human resource management. CGV's workforce reduction led to a decline in service quality and an increase in customer complaints in the short term.
There were more than 300 people waiting in line at the ticket booth, and the situation was such that even basic information services could not be provided. In the long term, the loss of experienced personnel resulted in the loss of operational know-how and additional costs associated with training new personnel.
As such, failure to predict demand has a devastating impact on a company's reputation and customer trust in the market. In addition, companies that fail to predict demand are unable to respond quickly to market changes, resulting in the loss of market share to competitors.
Failure to predict demand, it is important to analyse the structural causes
The root cause of demand forecasting failures lies in structural flaws that are a complex combination of institutional, technological, and organisational aspects. The most serious problem is the lack of political independence of forecasting agencies.
As seen in the case of the Singbu Line, local development logic and election pledges distort objective demand forecasts, and forecasting agencies present optimistic forecasts under the tacit pressure of local governments or the central government.
In this situation, even if a serious error occurs, such as the Yongin Light Rail Transit or the Uijeongbu Light Rail Transit, which is less than 30% of the forecast, there is no institutional mechanism to hold them accountable.
Ultimately, the uncertainty of responsibility is leading to the result of encouraging easy predictions.
There are also serious limitations in terms of technology. Current demand forecasting uses a rigid methodology that relies too much on historical data, making it very vulnerable to rapid changes in the environment or exceptional situations, as shown in the case of the movie theatre industry's response to COVID-19.
The semiconductor supply and demand failure in the automotive industry demonstrates the importance of data integration, as it shows that it is difficult to predict the complex interactions of the global supply chain with data from individual companies or industries alone.
On the organisational side, the lack of expertise and the lack of a system of interdepartmental cooperation are prominent. Many organisations rely on manual work using Excel, and there is a lack of specialists who can utilise advanced forecasting techniques.
Even more serious is the lack of a feedback system. Even in the case of the Singi-Bundang Line, a project that cost hundreds of billions of won, the cause of the failure to predict has not been properly analysed, which results in a missed opportunity for learning and improvement.
Since there is no post-evaluation, there is no systematic mechanism to improve the prediction system, and there is no virtuous cycle to learn from failed predictions and reflect them in the next prediction.
These structural problems are closely connected to each other, so it is difficult to solve them by improving only one factor. Political independence must be secured, advanced forecasting techniques introduced, professional personnel trained, interdepartmental cooperation strengthened, and a systematic post-evaluation system established.
What is particularly noteworthy is that these structural problems require fundamental changes in systems and organisational culture, rather than simply technological limitations or personnel issues. Therefore, the innovation of demand forecasting systems must go beyond the introduction of technology and lead to overall innovation of organisations and systems.
Innovation plan for demand forecasting systems
To improve the reliability of demand forecasting, a complete overhaul of the forecasting system is required. The recent emergence of AI-based demand forecasting systems is accelerating this innovation. Looking at the case of IMPACTIVE AI, AI-based demand forecasting systems are overcoming the limitations of traditional statistical methods.
Conventional demand forecasting often uses simple statistics such as the average of the past three months or manual Excel work. However, IMPACTIVE AI's predictive analytics system, Deepflow, automatically collects and analyzes 50,000 internal and external data to enable more sophisticated forecasting.
In particular, it increases the accuracy of forecasting by using various data such as ERP data, environmental data, and augmented/synthetic data.
The core of Deepflow, an AI-based demand forecasting system, is an ensemble approach that simultaneously utilises more than 200 predictive models. This is much more than the 10-20 models operated by a typical internal AI team at a company or the 80 models operated by a large IT company.
This multi-model approach allows for the consideration of various variables and situations, and contributes to improving the accuracy of predictions. In particular, it shows excellent prediction performance even in situations where there is a lack of historical data, such as new product launches. In fact, in new product demand forecasting, AI-based models have produced results that are more than twice as accurate as those of traditional statistical methods.
Future directions for demand forecasting
Demand forecasting systems are expected to evolve further in the future. In particular, the following developments are expected.
Real-time data integration analysis will become possible. Currently, forecasts are mainly based on historical data, but in the future, market data, consumer behaviour data, and economic indicators that occur in real time will be able to be reflected immediately.
Scenario-based prediction will be advanced. It will automatically generate scenarios based on changes in various variables and develop to the level of suggesting a response strategy for each scenario.
Autonomous learning ability will be enhanced. The ability to analyse the difference between the prediction and the actual result and continuously improve the prediction model based on this will be improved.
Innovation in demand forecasting is no longer an option, but a necessity. In the face of intensifying global competition and accelerating market changes, accurate demand forecasting is directly linked to the survival of a company.
In particular, the public sector urgently needs to build a forecasting system that is independent of political influence. To this end, it is necessary to actively consider the introduction of an AI-based forecasting system. In the private sector, it is necessary to shift away from the existing empirical forecasting method and move towards scientific forecasting based on data.
The failure of demand forecasting can have a direct impact on a company's financial health and national finances, beyond simply being a question of forecasting accuracy. As shown in the case of the Singbu Line, incorrect forecasting can lead to financial waste of hundreds of billions of won.
In conclusion, innovation in demand forecasting systems is a task that cannot be put off any longer. Advances in AI technology will be the key driver that enables this innovation. Governments and companies should actively embrace these technological advances and build a more efficient and accurate decision-making system.
.svg)
