Why Accurate Inventory Forecasting Matters More Than Ever in the Cosmetics Industry

on

Inventory forecasting has become an essential component of modern supply chain and retail management, particularly in industries where customer preferences change rapidly. Cosmetic retailers face continuous challenges in balancing product availability with inventory costs, as inaccurate demand estimation may lead to stock shortages, excessive inventory, expired products, and unnecessary financial losses. While many businesses still rely on intuition or market trends to estimate future demand, data-driven forecasting techniques provide a more systematic approach to inventory planning.

The article reviewed here investigates two widely used forecasting methods—the Least Squares method and the Double Moving Average technique—to determine which approach better predicts inventory demand for a popular cosmetic product. Rather than introducing a sophisticated artificial intelligence model, the study demonstrates how relatively simple statistical forecasting methods can improve inventory decision-making when applied appropriately. The findings offer useful insights for researchers, industrial engineers, retail managers, software developers, and practitioners interested in demand forecasting, production planning, inventory optimization, and digital business management.

Bibliographic Information

Item Information
Article Title Comparative Analysis of the Least Squares Method and Double Moving Average Technique for Forecasting Product Inventory
Authors Surfa Yondri, Dwiny Meidelfi, Tri Lestari, Fanny Sukma, and I.S. Mutia
Journal Teknomekanik
Volume & Issue Volume 7, Issue 1
Publication Year 2024
Pages 74–84
DOI https://doi.org/10.24036/teknomekanik.v7i1.29672
Publisher Universitas Negeri Padang
License Creative Commons Attribution 4.0 International (CC BY 4.0)

1. Research Background

  • Inventory management has become increasingly important in the cosmetics industry. The rapid growth of the global cosmetics market has intensified competition among retailers, making accurate inventory planning a critical factor for maintaining profitability, customer satisfaction, and operational efficiency. Cosmetic products often experience fluctuating demand due to fashion trends, product launches, seasonal preferences, and changing consumer behavior.
  • Traditional inventory decisions frequently depend on intuition and market trends. Many retailers estimate future demand by observing current market popularity rather than analyzing historical sales data systematically. Although this approach is simple to implement, it often produces unreliable purchasing decisions because consumer preferences may change quickly and unexpectedly.
  • Poor forecasting creates both financial and operational risks. Underestimating future demand can result in stock shortages, missed sales opportunities, and declining customer satisfaction. Conversely, excessive inventory increases storage costs, ties up working capital, and raises the risk of product obsolescence or expiration—particularly for cosmetic products with limited shelf life.
  • Data-driven forecasting offers a more objective alternative. Advances in inventory analytics have encouraged businesses to replace subjective estimation with quantitative forecasting techniques that utilize historical sales records. Such methods help organizations identify trends, estimate future demand, and improve inventory planning through measurable statistical models.
  • Numerous forecasting techniques are available, each with different assumptions. Previous studies have introduced approaches such as Single Moving Average, Weighted Moving Average, Double Moving Average, Single Exponential Smoothing, and Least Squares. Each method performs differently depending on whether sales data exhibit stable trends, seasonal variation, or short-term fluctuations.
  • Selecting an appropriate forecasting method remains a practical challenge. No forecasting model consistently outperforms all others across every business environment. The effectiveness of a forecasting technique depends largely on the characteristics of the available data, making comparative evaluation necessary before implementation in real operational settings.
  • The study addresses a practical research gap. Although inventory forecasting has been widely investigated, relatively few studies compare the performance of classical statistical forecasting techniques using real sales data from Indonesian cosmetic retailers while simultaneously integrating the methods into a decision-support system developed through the Software Development Life Cycle (SDLC).
  • The novelty lies in combining forecasting evaluation with system implementation. Rather than merely comparing mathematical forecasting accuracy, the authors develop a software application that implements the Least Squares and Double Moving Average methods, validates system functionality through software testing, and evaluates forecasting performance using Mean Absolute Percentage Error (MAPE). This integrated perspective strengthens the practical applicability of the research.

2. Research Objectives

  • To develop an inventory forecasting application capable of assisting inventory management for cosmetic retail businesses.
  • To implement the Least Squares forecasting method for predicting future inventory demand using historical sales data.
  • To implement the Double Moving Average forecasting technique with different moving-average ranges for inventory prediction.
  • To compare the forecasting performance of the Least Squares method and Double Moving Average technique using Mean Absolute Percentage Error (MAPE).
  • To identify the forecasting method that provides the most suitable prediction accuracy for Lip Cream Implora 01 inventory at Liza Cosmetics Shop.
  • To demonstrate how data-driven forecasting can support more effective inventory planning, reduce forecasting errors, and improve operational decision-making within the cosmetics retail sector.

3. Why This Research Matters

  • Supports evidence-based inventory management. The study demonstrates how statistical forecasting methods can replace subjective inventory estimation with objective, data-driven decision-making, enabling businesses to manage stock more systematically.
  • Improves operational efficiency. More accurate demand forecasting helps retailers maintain appropriate inventory levels, reducing unnecessary storage costs while minimizing the likelihood of product shortages.
  • Strengthens retail competitiveness. Businesses capable of predicting customer demand more accurately are better positioned to respond to market changes, improve service quality, and maintain customer satisfaction in highly competitive retail environments.
  • Demonstrates the practical value of classical forecasting methods. While recent research often emphasizes machine learning and artificial intelligence, this study shows that conventional statistical techniques remain valuable for organizations with limited computational resources and relatively straightforward sales patterns.
  • Bridges software engineering and business analytics. By incorporating forecasting algorithms into a software application developed using the SDLC Waterfall model, the research illustrates how analytical methods can be translated into practical decision-support systems suitable for everyday business operations.
  • Contributes to sustainable inventory management. Better forecasting reduces unnecessary inventory accumulation, minimizes product waste, improves resource utilization, and supports more sustainable retail operations by avoiding excessive purchasing.
  • Provides transferable insights beyond cosmetics. Although the case study focuses on a cosmetic retailer, the forecasting principles discussed in the article can also be applied to pharmacies, supermarkets, fashion retailers, food distribution, spare-part inventory, manufacturing supply chains, and other industries where demand prediction plays an essential role in inventory planning.

4. Research Methodology

  • Research Type

    This study employed quantitative applied research with a case study approach focusing on inventory demand forecasting in the cosmetics retail sector. The research combined statistical forecasting techniques with software development to produce a practical decision-support application for inventory management rather than solely evaluating mathematical forecasting performance.

  • Case Study

    The investigation used historical sales data of Lip Cream Implora 01 from Liza Cosmetics Shop covering the period from April 2021 to June 2022. This product was selected because its sales pattern represented the inventory challenges commonly encountered in cosmetic retailing, including demand fluctuations and changing purchasing behavior.

  • Research Framework

    Instead of directly comparing forecasting equations, the researchers developed a complete forecasting application following the Software Development Life Cycle (SDLC) using the Waterfall development model. This structured framework ensured that forecasting algorithms were systematically translated into an operational software system capable of supporting inventory planning.

  • Software Development Model

    The Waterfall model consisted of six sequential phases:

    • Literature review and identification of forecasting approaches.
    • Requirement analysis and data collection.
    • System and software design using UML, interface design, and database planning.
    • Implementation of forecasting algorithms together with unit testing.
    • System integration and functional validation using black-box testing.
    • Deployment and maintenance.

    The workflow demonstrates that forecasting accuracy alone was not the only objective; software reliability and usability were also considered during system development.

  • Forecasting Models

    Two classical forecasting techniques were implemented and compared:

    • Least Squares Method, which estimates a linear relationship between time and historical sales to predict future demand.
    • Double Moving Average (DMA), evaluated using both three-period and five-period moving average windows to accommodate short-term demand fluctuations.

    The comparison enabled the researchers to evaluate how different statistical assumptions influenced forecasting performance under identical sales data.

  • Forecast Accuracy Evaluation

    Forecasting performance was evaluated using Mean Absolute Percentage Error (MAPE), one of the most widely accepted measures for forecasting accuracy. The study classified prediction quality according to predefined MAPE categories ranging from very good to poor, allowing straightforward interpretation of each method's effectiveness.

  • Software Validation

    The developed application underwent two complementary validation procedures. White-box testing verified the correctness of the implemented program logic and source code, whereas black-box testing examined system functionality from the user's perspective by evaluating inputs, outputs, and boundary conditions.

  • Comparative Analysis

    The final stage compared forecasting results produced by all implemented methods using identical historical sales data. Prediction accuracy, forecast values, and corresponding MAPE scores were analyzed to identify the most appropriate technique for inventory forecasting within the selected case study.

5. Key Findings

Least Squares Produced the Most Accurate Forecast

Among the forecasting techniques evaluated, the Least Squares method achieved the best overall performance. Using historical sales data collected over fifteen months, the model generated a demand prediction of 13 units for the subsequent period while producing the lowest Mean Absolute Percentage Error (MAPE) among all tested approaches.

Although the reported MAPE of 48% is categorized as "sufficient" rather than excellent, it still substantially outperformed both Double Moving Average configurations. This finding suggests that the historical sales pattern exhibited a relatively stable long-term trend that could be represented more effectively through linear regression than through moving-average smoothing.

Double Moving Average Showed Lower Forecasting Accuracy

The Double Moving Average technique demonstrated weaker predictive performance under the selected dataset. Using a three-period moving average, the model forecasted demand of 27 units but produced a MAPE of 94%, indicating poor forecasting accuracy according to the study's evaluation criteria.

Similarly, the five-period configuration generated an even more conservative inventory recommendation while still producing a relatively high forecasting error. These outcomes indicate that the moving-average approach struggled to capture the underlying sales pattern observed in the case study, particularly when sales fluctuated substantially across different months.

Historical Sales Characteristics Influence Method Selection

An important conclusion emerging from the comparison is that forecasting performance depends heavily on the characteristics of historical sales data rather than on the popularity of a forecasting technique. Methods that perform well under one demand pattern may not necessarily provide similar accuracy for another product or business environment.

The study illustrates that the Least Squares approach was better suited for datasets exhibiting relatively consistent long-term movement, whereas Double Moving Average methods may be more appropriate under different demand dynamics. Consequently, organizations should evaluate forecasting techniques using their own operational data before adopting a particular model.

Software-Based Forecasting Improves Inventory Decision Support

Beyond comparing forecasting equations, the research demonstrates the practical feasibility of embedding forecasting algorithms within an inventory management application. The developed software allows forecasting calculations to be performed systematically rather than manually, reducing calculation errors while improving decision consistency.

The successful implementation of both forecasting methods within the developed application illustrates how relatively simple statistical models can be transformed into practical business tools that assist inventory managers in planning future purchasing decisions.

Accurate Forecasting Can Reduce Inventory Risk

The findings reinforce the importance of accurate forecasting as an integral component of inventory management. Better demand prediction enables businesses to reduce the likelihood of overstocking and stock shortages while improving resource allocation throughout purchasing and storage operations.

For cosmetic retailers, where consumer preferences evolve rapidly and many products have limited shelf lives, improved forecasting contributes directly to lower operational costs, better customer satisfaction, and more efficient inventory turnover.

Classical Statistical Methods Remain Valuable in Practice

Although recent research frequently focuses on artificial intelligence and machine learning, this study demonstrates that established statistical forecasting methods continue to provide meaningful value for practical business applications. Organizations with limited computational resources can still improve inventory planning by selecting forecasting methods that appropriately match their historical sales characteristics.

The research therefore highlights that successful forecasting depends not only on algorithmic complexity but also on understanding data characteristics, selecting suitable analytical techniques, and integrating them into operational business processes.

6. Scientific Contribution

  • Provides an empirical comparison between the Least Squares method and Double Moving Average techniques using real inventory data from the cosmetics retail industry.
  • Demonstrates the practical implementation of statistical forecasting methods within an operational software application developed using the SDLC Waterfall model.
  • Shows that forecasting accuracy depends on data characteristics rather than assuming that one forecasting method is universally superior.
  • Combines software engineering with inventory analytics, illustrating how forecasting models can be integrated into decision-support systems for business operations.
  • Contributes practical evidence supporting the continued relevance of classical statistical forecasting techniques in small and medium-sized retail businesses.
  • Provides a reproducible evaluation framework based on MAPE together with software validation through white-box and black-box testing.

7. Industrial Implications

  • Improves inventory planning. Retail businesses can utilize demand forecasting to determine purchasing quantities more accurately, reducing unnecessary inventory accumulation.
  • Reduces operational costs. Better forecasting minimizes expenses related to overstocking, warehousing, expired products, and emergency purchasing caused by inventory shortages.
  • Supports production and supply planning. Manufacturers and distributors can align production schedules and replenishment activities with more reliable demand estimates.
  • Enhances customer satisfaction. Maintaining appropriate inventory availability reduces stockouts, ensuring that customers can obtain products when required.
  • Strengthens digital inventory management. The integration of forecasting algorithms into software applications encourages organizations to adopt digital decision-support systems rather than relying on manual calculations.
  • Supports Industry 4.0 transformation. Although the forecasting methods are relatively simple, embedding analytical models within software platforms represents an important step toward digital business operations and data-driven management.
  • Applicable across multiple industries. Similar forecasting approaches may also support inventory management in pharmacies, supermarkets, manufacturing warehouses, spare-parts distribution, healthcare logistics, food retailing, and e-commerce fulfillment.

8. Research Limitations

  • Single-product case study. The investigation focused exclusively on one cosmetic product (Lip Cream Implora 01) from a single retail store. Consequently, the findings primarily reflect the sales characteristics of this particular product and business context rather than representing the broader cosmetics industry.
  • Limited historical dataset. The forecasting models were developed using historical sales records collected from April 2021 to June 2022. Although sufficient for comparing the selected forecasting methods, a longer observation period could reveal additional demand patterns, including long-term trends and cyclic behavior.
  • Restricted forecasting techniques. The study evaluated only the Least Squares method together with Double Moving Average using two moving-average ranges. Other forecasting approaches, such as exponential smoothing, ARIMA, machine learning, or hybrid forecasting models, were outside the scope of this research.
  • Forecasting relied primarily on historical sales data. External variables that frequently influence cosmetic demand—such as promotional campaigns, social media trends, seasonal events, competitor activities, and macroeconomic conditions—were not incorporated into the forecasting models.
  • Software validation emphasized functionality. The developed application successfully passed white-box and black-box testing; however, the research did not include long-term field implementation to evaluate user acceptance, operational performance, or organizational impacts after deployment.
  • Prediction accuracy remained moderate. Although the Least Squares method outperformed the Double Moving Average approaches, its reported Mean Absolute Percentage Error (MAPE) still indicates that forecasting uncertainty remains present. The results therefore demonstrate relative improvement rather than perfect prediction accuracy.

9. Future Research Opportunities

  • Evaluate additional statistical and machine learning forecasting models, including ARIMA, Holt-Winters, Random Forest, Support Vector Regression, Long Short-Term Memory (LSTM), and hybrid forecasting approaches.
  • Extend the analysis to multiple cosmetic products representing different demand characteristics, product life cycles, and customer purchasing patterns.
  • Investigate forecasting performance using larger datasets covering multiple years to capture seasonal effects, long-term trends, and structural market changes.
  • Integrate external variables—including promotional campaigns, holidays, inflation, social media influence, weather conditions, and customer demographics—into forecasting models to improve predictive performance.
  • Develop intelligent inventory management systems capable of automatically recommending purchasing quantities based on continuously updated forecasting results.
  • Implement real-time forecasting systems connected to point-of-sale (POS) databases and cloud-based inventory platforms to support continuous inventory monitoring.
  • Compare forecasting performance across different retail sectors, including pharmacies, supermarkets, fashion retailers, electronic products, and manufacturing spare-parts inventory.
  • Conduct economic impact analyses to quantify the financial benefits of improved forecasting accuracy in terms of inventory cost reduction, cash-flow improvement, and return on investment.
  • Investigate user acceptance and organizational readiness for adopting forecasting-based decision-support systems within small and medium-sized enterprises.
  • Explore the integration of forecasting applications with Enterprise Resource Planning (ERP), Supply Chain Management (SCM), Business Intelligence (BI), and Industry 4.0 digital manufacturing ecosystems.

10. Potential for Public Policy Citation (Overton)

The direct policy relevance of this article is relatively moderate because its primary contribution concerns forecasting methodology and inventory management within a single retail case study rather than national economic policy or regulatory development. Nevertheless, the research demonstrates how data-driven inventory forecasting can improve operational efficiency, reduce business waste, and strengthen digital decision-making for retail enterprises.

The study could therefore provide supporting evidence for government initiatives promoting digital transformation among small and medium-sized enterprises (SMEs), retail digitalization strategies, and programs encouraging data-driven business management. Agencies responsible for digital economy development, entrepreneurship support, or industrial modernization may find the study relevant as an illustrative example of practical forecasting implementation.

The article is less likely to be cited directly in engineering standards, manufacturing regulations, or national inventory policies because its scope remains limited to methodological comparison and software implementation. Broader policy influence would require additional validation involving multiple businesses, larger datasets, and measurable economic impacts across different industrial sectors.

11. Who Should Read This Paper?

  • Researchers working in forecasting, operations research, and inventory management.
  • Industrial engineers involved in production planning and supply chain optimization.
  • Supply chain and logistics professionals responsible for inventory control.
  • Retail managers seeking evidence-based inventory planning techniques.
  • Software developers building business decision-support systems.
  • Business analysts interested in forecasting and operational analytics.
  • Graduate students studying industrial engineering, information systems, operations management, or data analytics.
  • Small and medium-sized enterprise (SME) owners aiming to improve inventory performance.
  • Educators teaching forecasting methods, business analytics, or production management.
  • Government agencies promoting digital transformation and data-driven business practices.

12. Final Thoughts

This article presents a practical and well-structured comparison of two widely used statistical forecasting techniques within the context of cosmetic retail inventory management. Rather than pursuing increasingly complex predictive models, the authors demonstrate that selecting an appropriate forecasting method based on historical sales characteristics can significantly improve inventory planning. Their comparison between the Least Squares method and the Double Moving Average technique illustrates that forecasting accuracy depends not only on the mathematical model itself but also on how well the selected approach matches the underlying demand pattern.

One of the study's major strengths lies in its integration of forecasting methodology with software engineering. By implementing both forecasting techniques within an application developed through the Software Development Life Cycle (SDLC), the research moves beyond theoretical comparison and demonstrates how forecasting algorithms can become practical decision-support tools for everyday business operations. This combination of analytical modelling, software implementation, and functional validation increases the practical value of the work for both academic researchers and industry practitioners.

Although the case study focuses on a single cosmetic product, the principles discussed throughout the article extend well beyond the cosmetics industry. Similar forecasting challenges exist across manufacturing, retail, healthcare logistics, food distribution, warehouse management, and e-commerce operations. Consequently, the study provides a useful reference for organizations seeking to improve inventory decisions through accessible statistical forecasting methods without requiring sophisticated artificial intelligence infrastructure.

Overall, this research contributes meaningful practical knowledge to inventory forecasting by demonstrating that relatively simple quantitative methods can support better operational decisions when appropriately selected and systematically implemented. It also establishes a solid foundation for future studies that integrate traditional forecasting techniques with modern analytics, digital inventory platforms, and intelligent decision-support systems.

Suggested Citation

Teknomekanik (UNP) Style

Yondri, S., Meidelfi, D., Lestari, T., Sukma, F., & Mutia, I.S. (2024). Comparative analysis of the least squares method and double moving average technique for forecasting product inventory. Teknomekanik, 7(1), 74–84. https://doi.org/10.24036/teknomekanik.v7i1.29672

APA (7th Edition)

Yondri, S., Meidelfi, D., Lestari, T., Sukma, F., & Mutia, I. S. (2024). Comparative analysis of the least squares method and double moving average technique for forecasting product inventory. Teknomekanik, 7(1), 74–84. https://doi.org/10.24036/teknomekanik.v7i1.29672

IEEE Style

S. Yondri, D. Meidelfi, T. Lestari, F. Sukma, and I. S. Mutia, "Comparative analysis of the least squares method and double moving average technique for forecasting product inventory," Teknomekanik, vol. 7, no. 1, pp. 74–84, Jun. 2024, doi:10.24036/teknomekanik.v7i1.29672.

Harvard Style

Yondri, S., Meidelfi, D., Lestari, T., Sukma, F. & Mutia, I.S., 2024. Comparative analysis of the least squares method and double moving average technique for forecasting product inventory. Teknomekanik, 7(1), pp.74–84. Available at: https://doi.org/10.24036/teknomekanik.v7i1.29672.

Vancouver Style

Yondri S, Meidelfi D, Lestari T, Sukma F, Mutia IS. Comparative analysis of the least squares method and double moving average technique for forecasting product inventory. Teknomekanik. 2024;7(1):74-84. doi:10.24036/teknomekanik.v7i1.29672.

Chicago (Author–Date)

Yondri, Surfa, Dwiny Meidelfi, Tri Lestari, Fanny Sukma, and I. S. Mutia. 2024. "Comparative Analysis of the Least Squares Method and Double Moving Average Technique for Forecasting Product Inventory." Teknomekanik 7 (1): 74–84. https://doi.org/10.24036/teknomekanik.v7i1.29672.

MLA (9th Edition)

Yondri, Surfa, et al. "Comparative Analysis of the Least Squares Method and Double Moving Average Technique for Forecasting Product Inventory." Teknomekanik, vol. 7, no. 1, 2024, pp. 74–84. Crossref, https://doi.org/10.24036/teknomekanik.v7i1.29672.

Editorial Note

Editorial Note: This blog post is an independent scholarly review intended for educational and scientific communication purposes. It summarizes and discusses the published article in the author's own words while providing full attribution to the original publication, consistent with the principles of the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

Engineering Research Insights
Advancing Engineering Knowledge Through Scholarly Communication

Comments

Post a Comment