Demand Forecasting in the Presence of Systematic Events: Cases in Capturing Sales Promotions

Mahdi Abolghasemi,Ali Eshragh,Jason Hurley,Behnam Fahimnia

DOI: https://doi.org/10.48550/arXiv.1909.02716

2019-09-06

Abstract:Reliable demand forecasts are critical for the effective supply chain management. Several endogenous and exogenous variables can influence the dynamics of demand, and hence a single statistical model that only consists of historical sales data is often insufficient to produce accurate forecasts. In practice, the forecasts generated by baseline statistical models are often judgmentally adjusted by forecasters to incorporate factors and information that are not incorporated in the baseline models. There are however systematic events whose effect can be effectively quantified and modeled to help minimize human intervention in adjusting the baseline forecasts. In this paper, we develop and test a novel regime-switching approach to quantify systematic information/events and objectively incorporate them into the baseline statistical model. Our simple yet practical and effective model can help limit forecast adjustments to only focus on the impact of less systematic events such as sudden climate change or dynamic market activities. The proposed model and approach is validated empirically using sales and promotional data from two Australian companies. Discussions focus on a thorough analysis of the forecasting and benchmarking results. Our analysis indicates that the proposed model can successfully improve the forecast accuracy when compared to the current industry practice which heavily relies on human judgment to factor in all types of information/events.

Applications

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to improve the accuracy of demand forecasting in the presence of systemic events (such as sales promotions). Specifically, traditional statistical models based on historical sales data often fail to fully capture the impact of these systemic events, resulting in less accurate forecasts. Therefore, forecasters usually need to adjust these baseline forecasts based on experience and judgment. However, such adjustments are often subjective and unsystematic, which may lead to forecast bias.

To solve this problem, the author proposes a new **regime - switching approach** to quantify and model the impact of systemic events and objectively incorporate them into the baseline statistical model. In this way, the need for manual adjustment can be reduced, making the forecast more accurate and reliable.

### Main problem summary:
1. **Limitations of traditional models**: Statistical models relying solely on historical sales data cannot fully consider the impact of systemic events (such as sales promotions), resulting in inaccurate forecasts.
2. **Inadequacy of manual adjustment**: In current practice, forecasters rely on experience and judgment to adjust baseline forecasts, but this method lacks systematicness and consistency.
3. **Quantification of systemic events**: How to effectively quantify and model the impact of systemic events (such as sales promotions) to improve the accuracy of forecasting.

### Solution:
The author has developed a time - series regression model based on **regime - switching**, called the **Forecasting Systematic Events (FSE) model**. This model improves demand forecasting in the following ways:
- **Identifying systemic events**: Determine the most significant systemic events and their impact levels through statistical analysis (such as Analysis of Variance, ANOVA).
- **Defining demand uplift states**: Define different Demand Uplift States (DUS) according to different combinations of systemic events.
- **Embedding in the model**: Embed these demand uplift states into the time - series regression model to capture the impact of systemic events on demand.

Through this method, the FSE model can more accurately predict demand without relying on a large amount of manual intervention, especially in the presence of systemic events such as sales promotions.

### Formula representation:
The specific formula of the FSE model is as follows:

\[
X_t=\alpha_0+\sum_{i = 1}^{p}\alpha_iX_{t - i}+\sum_{j = 1}^{m}\beta_jS_{t,j}+\epsilon_t
\]

where:
- \(X_t\) is the demand quantity at time \(t\);
- \(S_{t,j}\) is the demand uplift state variable at time \(t\). If the demand uplift is in state \(j\), it takes the value of 1, otherwise 0;
- \(\epsilon_t\) is a Gaussian white - noise process;
- \(\alpha_i\) and \(\beta_j\) are unknown parameters, estimated through time - series data and DUS algorithm output;
- \(p\) is the number of past demand values regressed in the model;
- \(m\) is the total number of demand uplift states determined by the DUS algorithm.

When there are no systemic events, all state variables \(S_{t,j}\) are 0, and at this time the FSE model is simplified to an autoregressive model (AR model).