An Integrated Lean and Six Sigma Framework for Improving Productivity Performance: A Case Study in a Spanish Chemicals Manufacturer

Abstract

In the pursuit of operational excellence and enhanced competitiveness, a wide range of industries have turned to methodologies such as Lean and Six Sigma; however, in the chemical sector, their application is very limited. This paper presents a Lean Six Sigma framework to identify and reduce sources of variability occurring in the final product composition of a Spanish SME fertilizer manufacturer. The company faced important challenges related to product variability, adversely affecting overall productivity. A real-life case of the Lean Six Sigma implementation was conducted over two years, and its applicability and ability to improve productivity performance were thoroughly assessed. The proposed framework successfully integrated Lean and Six Sigma methodologies, i.e., process mapping (value stream mapping), root cause analysis (Ishikawa cause–effect diagram), project management (SIPOC and DMAIC), and statistical process control, and demonstrated practical benefits for the case company by identifying the key variables affecting product variability and determining their optimal levels. A substantial 50% reduction in the variability of several products and a 42% reduction in material preparation time were achieved. These reductions resulted in a 40% reduction in costs associated with product losses and a 54% reduction in costs from raw material losses.

1. Introduction

Spain is the tenth largest agri-food power in the world and the fourth largest agricultural area within the European Union [1]. In recent decades, the use of fertilizers has played a crucial role due to their great contribution to intensifying agricultural production in response to the increasing demand for food products by the growing global population [2]. In this context, the Spanish fertilizer industry has experienced a growing demand for its products, facing a continuing challenge to enhance production, efficiency, and product quality [3]. The manufacture of chemical fertilizers is subject to stringent regulatory standards that require consistent compositions of specified elements across all batches. The raw materials can be either solid or liquid, and the uniformity in their composition is crucial to ensure consistency in the final product content. However, achieving consistent product quality is often challenging due to the inherent variability in raw materials, manufacturing processes, and environmental conditions. Variability in product composition can have significant implications for both manufacturers and customers [4]. For manufacturers, it can result in increased production costs, rework, and waste, while for customers, it can lead to inconsistent product performance and reduced satisfaction [5]. Thus, maintaining the process within tolerance limits is essential to ensure consumer satisfaction and to thrive in a competitive market.

A potential strategy that can be used to minimize process variation is the integration of Lean and Six Sigma (LSS) methodologies [6]. LSS is a data-driven approach that combines the fundamental principles of Lean and Six Sigma to assist an organization in maintaining a competitive position in the global market and achieving business goals and organizational excellence [7]. In short, Lean focuses on waste elimination and speed reduction, improving process efficiency and reducing overall costs, with a particular emphasis on customer satisfaction [8]. It provides a set of tools and techniques designed to minimize lead times, inventories, setup times, rework, and other hidden factory inefficiencies [9]. On the other hand, Six Sigma aims to reduce variations and defects through statistical analysis and rigorous problem-solving tools [10]. It is a project-oriented, statistically based methodology that provides data to drive solutions. Hence, each approach offers a unique set of tools, and their integration can maximize the potential that either approach could achieve independently. Large enterprises, such as Toyota, Texas Instruments, Kodak, General Electric, Bank of America, and Sony, have achieved impressive results because of the adoption of these methodologies [11,12,13].

In chemical industries, such as fertilizer manufacturers, processes involve multiple variables, including large volumes of hazardous substances, strict regulations, energy consumption, and waste production. Consequently, this sector is suitable to fully benefit from these business strategies. This work presents a case study conducted in a Spanish chemical company focused on implementing an LSS framework in order to reduce variability in the final product composition. Three specific fertilizers exhibiting high composition variability were selected for analysis. A range of LSS tools, including process mapping (value stream mapping), root cause analysis (Ishikawa cause–effect diagram), project management (SIPOC and DMAIC), and statistical process control, was implemented in the company to monitor the impact of variability in production and its effect on inventory. Variability in product composition is a major factor in customer dissatisfaction and may jeopardize customer loyalty [14]. By minimizing it, companies can enhance not only customer satisfaction but also product quality and profits.

Literature Review

Over the past two decades, numerous companies, particularly in the manufacturing sector, have faced increased pressure from customers and competitors to deliver high-quality products at lower prices [15,16]. This pressure has mobilized manufacturing companies to adopt business strategies that improve their competitiveness and meet customer expectations while maximizing profit. Lean Six Sigma has been successfully implemented in a wide range of sectors, including financial and banking services [17], education [18], healthcare [19], IT [20], and wood [21]; however, its implementation in chemical companies is scarce. According to Alarcón et al. [22], the chemical sector only accounts for 4% of the manufacturing companies that have adopted Lean and Six Sigma models. Nonetheless, the effectiveness of these methodologies has been notably proven in this industry. For instance, Muganyi et al. [23] demonstrated the effectiveness of LSS as a strategic tool for ensuring business survival in a South African multinational chemical manufacturing company. The implementation of several LSS projects yielded dramatic results, including an 11% increase in production rate, a 9% reduction in cycle time, a five-fold decrease in claims, and a 15% reduction in overall cost per unit. Dow Chemical is another well-documented success case. This chemical multinational enterprise has registered impressive savings of more than USD 1.5 billion due to its ongoing dedication to these business improvement strategies [24]. For instance, substantial improvements in inventory management were achieved, leading to a 10% reduction in the average daily sales of agricultural chemicals stock in 2004. The productivity improvement and its ability to reduce costs can be attributed to the fact that nearly 60% of employers are exposed to Lean and Six Sigma principles. Du Pont is another large chemical company that has successfully adopted and applied Six Sigma for more than two decades. Du Pont integrated Six Sigma tools at the enterprise-wide level, including sales, marketing, HR, finance, and production. They carried out over 3000 Six Sigma tasks and provided training to more than 10,000 specialists, achieving cost savings of more than USD 1.5 billion over 10 years [25]. Lean Six Sigma was also applied to the supply chain of a leading fertilizer company in Turkey to optimize logistics operations, resulting in a reduction in inventory time from 82 to 51 days and a 36% decrease in the average inventory amount [26].

Many successful cases of LSS adoption are associated with large multinational corporations. In fact, these business models have been questioned due to the perceived requirement for significant investment and resource-intensive programs that only large companies could afford [27]. However, these methodologies can also be effectively applied in small and medium-sized enterprises (SMEs). Indeed, they may have even fewer complications in terms of project simplicity and the effort required for team building and employee training [28]. For example, NutriSoil, a Portuguese SME that sells fertilizer in bags, had difficulties with the filling process due to overweight bags. Implementing LSS, NutriSoil enhanced its Cpk index (capability index) in the filling process, which improved consumer satisfaction and cost savings [4]. LSS was also implemented to mitigate delays in delivering reports within a chemical laboratory [29]. The study showed that delays were caused by retests, which could be reduced by 5%, saving at least USD 200 for every 40 tests. Furthermore, Ishak et al. [30] used the DIMAC approach to enhance wastewater management in a Malaysian SME poultry plant and reported a 15.3% monthly reduction in chemical consumption. In another study, Wegner [31] applied waste reduction techniques in the analysis process of custom-blend fertilizer performed in an agricultural laboratory. This resulted in a substantial reduction in sample turnaround time, decreasing it from 45 to 3 days. This achievement had a positive impact on customer satisfaction and operational efficiency.

As can be seen, the adoption of LSS in the chemical industry has shown very promising results. However, the literature highlights a lack of research on implementing these methodologies in the chemical sector, especially in manufacturing fertilizers. In particular, no studies have been found dealing with the implementation of Lean and Six Sigma tools to minimize product composition variability in a fertilizer company. Furthermore, Spanish fertilizer companies also lack the adoption and effective implementation of LSS practices. The limited or non-existent implementation of these methodologies in Spanish fertilizer companies also contributes to the lack of industry collaboration and the limited knowledge of LSS in this specific field (e.g., success factors, guidelines, tools, etc.). Therefore, further research is needed to explore and identify the barriers and factors that contribute to the successful implementation of LSS in this specific context. In this regard, this study presents a real-life case of LSS adoption in a Spanish SME fertilizer company to assess its effectiveness in addressing a complex operational challenge.

2. Materials and Methods

A case study is a research method that involves in-depth investigation and analysis of a particular subject, phenomenon, or entity within its real-life context [32]. It typically involves gathering detailed qualitative and/or quantitative data from multiple sources such as interviews, observations, documents, and experimental tests, offering practical insights and solutions based on empirical evidence. The present case study was conducted in a Spanish chemical company specializing in fertilizer manufacturing. The company faced challenges with product variability, which detrimentally affected overall productivity. Thus, this research aimed to investigate the variability in the compositions of three chemical products and evaluate the application of an LSS framework to improve operational efficiency and boost company productivity. The targeted products are marketed in liquid form and packaged according to their specific weight. Characteristic values for each product were monitored, i.e., density and main ingredient compositions. For confidentiality reasons, the specific products are referred to as P₁, P₂, and P₃. In P₁, the studied component is supplied in solid form. In P₂, two components were monitored, both provided in liquid form. In P₃, the analyzed component is also provided in liquid form but present in a higher proportion. Figure 1 shows the methodology followed for this study.

Firstly, the problem statement was defined, and the dependent and independent variables were identified. Subsequently, data were collected and analyzed over a two-year period, determining the potential losses generated during the manufacturing processes and identifying areas for process improvement. Then, a problem-solving approach based on LSS tools was designed and implemented in the company. The following tools were selected: Ishikawa diagram, SIPOC (Suppliers, Input, Process, Outputs, and Customers) diagram, value stream map (VSM), descriptive statistics (regression analysis, normal distribution, etc.), and DMAIC (Define, Measure, Analyze, Improve, and Control). Finally, post-implementation data were collected, and key findings were analyzed and elucidated.

2.1. Lean Tools Deployment

2.1.1. Ishikawa Diagram

The Ishikawa diagram, also known as the cause–effect diagram, was used to identify the potential factors causing variability in the main components of the products. The causes or influencing factors were categorized into main groups to facilitate the identification of the sources of variation [33]. Furthermore, several items were placed as backbone spines, denoting the root causes. Figure 2 shows the fishbone diagram designed to determine the main causes contributing to alterations in the composition of the products under study.

The most important factors affecting the product variability were grouped into six categories following the 6M method [33]: methods, environment, facilities, materials, staff, and measurement. From this division, several root causes were identified: production haste or insufficient time allocation, inefficient work instructions, cleanliness and distribution of work areas, machinery availability and proper calibration, variability in the compositions of raw materials, adequate training and precision of involved personnel, inaccurate measurements, and fluctuations in raw material compositions.

2.1.2. SIPOC Diagram

The entire process was evaluated using SIPOC, i.e., Suppliers, Input, Process, Outputs, and Customers, a powerful six-sigma mapping tool that helps optimize operations and enhance customer satisfaction [34]. This step further refines the problem identification phase, highlighting the direct relationship between customers and suppliers and interrelating the key processes within an organization. In this study, the SIPOC diagram was used to define the initial phase of the DIMAC process. Figure 3 shows the SIPOC diagram applied to the product development under investigation.

In chemical manufacturing, product variability is often attributed to factors such as raw material variability, process-related factors, and challenges with measurement and monitoring systems. Thus, alterations in product composition are likely to occur at specific points within stage P (process), suggesting that process improvement efforts should be focused on this stage. Stage P mainly involves the generation of standard templates or recipes for product preparation and encompasses essential production steps, i.e., raw material selection, transportation, weighing, dosing, performing reaction, etc. Therefore, these steps were subjected to a thorough analysis in order to develop an effective strategy to achieve a reduction in variability in product composition.

2.1.3. Value Stream Map

In any production process, there are specific stages that are crucial for achieving the company’s goals. The failure of these stages can create a bottleneck in the production line, leading to delays, increased costs, and decreased overall efficiency. Hence, it is essential to identify and exert control over these production stages. The value stream map (VSM) is a Lean tool employed for this purpose. It facilitates the visualization and analysis of the entire production process, allowing for the identification of waste areas and potential avenues for improvement. Figure 4 shows the VSM designed for this study, including the timeline of each stage and the lead time ladder. Time data were collected and analyzed using production records from a two-year period, providing average values for each phase of the process.

According to the VSM, the greatest number of incidents related to the inappropriate use of materials could occur during stages directly involved in the handling of raw materials. Furthermore, the total duration of these stages is the longest of the entire process, i.e., 3.5 h, excluding logistics. Consequently, the LSS application framework was focused on the following stages: raw material preparation, weighing, and dosing. Finally, a problem-solving approach, i.e., DMAIC, was applied (

Table 1. Application of the DMAIC cycle to the present work.

Phase	Contents	Application
Define	Indicate the main objective of the application, as well as the critical project to be developed	Reducing possible losses of ingredients used in manufacturing. Associated with this are also the purchasing process, the generation of waste, and the use of energy resources
Measure	Develop a data collection plan and compare data to identify problems and gaps	Analysis of manufactured products. Relationship with the composition of raw materials and the reliability of added quantities
Analyze	Determine the causes of defects and the sources of variation	Establish relationships between the data obtained through analysis using graphical techniques, statistics, etc.
Improve	Propose measures to reduce or eliminate variations	Study trends and possible corrective measures to be implemented
Control	Check that the process variations comply with the established requirements	Determine the impact of the proposed changes on the processes and their follow-up

).

These steps were followed for the successful implementation of the proposed LSS framework in the case company. Additionally, statistical tools were applied to analyze the data and identify patterns or trends.

3. Results and Discussion

3.1. Define

Each product (P_i) is composed of several ingredients, and each ingredient (m_j) has a fixed composition in at least one of the components required to manufacture the product. Generally, it can be assumed that each ingredient (m_j) contributes only one component. Thus, the mass C_ij (kg) of each component can be obtained using the mass fractions (X_ij):

$$C_{11}=X_{11}·m_1$$

(1)

$$C_{12}=X_{12}·m_2$$

(2)

$$C_{13}=X_{13}·m_3$$

(3)

$$C_{ij}=X_{ij}·m_j$$

(4)

Product P_i can be expressed as a combination of its ingredients (m_j), considering they represent a proportion of the total (Y_ij):

$$P_i=m_1+m_2+\cdots +m_j$$

(5)

$${\displaystyle \frac{1}{X_{i1}}}·{\displaystyle \frac{C_{i1}}{P_i}}+{\displaystyle \frac{1}{X_{i2}}}·{\displaystyle \frac{C_{i2}}{P_i}}+\cdots +{\displaystyle \frac{1}{X_{ij}}}·{\displaystyle \frac{C_{ij}}{P_i}}=1$$

(6)

$$Y_{ij}={\displaystyle \frac{1}{X_{ij}}}·{\displaystyle \frac{C_{ij}}{P_i}}$$

(7)

The target values (T) for each component j in product i can be expressed as follows:

$$Y_{i1T}, Y_{i2T}, \dots , Y_{ijT}$$

(8)

and the corresponding actual values (R) for the content of each component are

$$Y_{i1R}, \dots , Y_{ijR}$$

(9)

The objective was to minimize the differences between the real and the theoretical values, with the ideal situation being

$$Y_{ijT}-Y_{ijR}=0$$

(10)

Using the information collected from the company production plan, a total of 35 batches of each product were produced over the course of one year. The density and the composition of the key ingredients were monitored and compared to the theoretical values provided by the supplier. The differences between the actual (Y_ijR) and the theoretical (Y_ijT) values indicate variability in product composition, which in turn reflects potential product losses. For each product, the following characteristic components were studied: Y₁₁ for product P₁; Y₂₁ and Y₂₂ for product P₂ (it has two components); and Y₃₁ for product P₃. Since the variation in density (D_i) is directly correlated with potential material loss, the differences between the theoretical density values and those obtained in each batch (D_iT − D_iR) were also monitored.

3.2. Measure

The densities and compositions of the main ingredients were assessed across 35 productions of P₁, P₂, and P₃ products. Densities were measured using a rotary viscosimeter densimeter, and compositions were determined with an inductively coupled plasma mass spectrometer.

Table 2. Data collection for P₁; theoretical values: D_1T = 1.200 kg/L; Y_11T = 0.055 kg C_11T/kg T.

Test	D1R kg/L	Y11R kg C11R/kg T	Y11T − Y11R kg C11/kg T	D1T − D1R kg/L
1	1.222	0.057	−0.002	−0.029
2	1.222	0.057	−0.002	−0.027
3	1.226	0.051	0.004	−0.019
4	1.229	0.055	0.000	−0.008
5	1.227	0.053	0.002	−0.013
6	1.219	0.054	0.001	−0.029
7	1.208	0.047	0.008	−0.010
8	1.213	0.052	0.003	−0.026
9	1.229	0.051	0.004	−0.026
10	1.210	0.051	0.004	−0.020
11	1.226	0.057	−0.002	−0.027
12	1.226	0.056	−0.001	−0.030
13	1.220	0.057	−0.002	−0.028
14	1.227	0.052	0.003	−0.025
15	1.230	0.052	0.003	−0.012
16	1.228	0.052	0.003	−0.014
17	1.225	0.055	0.000	−0.030
18	1.212	0.055	0.000	−0.032
19	1.214	0.052	0.003	−0.017
20	1.230	0.051	0.004	−0.026
21	1.232	0.051	0.004	−0.033
22	1.217	0.054	0.001	−0.012
23	1.226	0.052	0.003	−0.012
24	1.233	0.057	−0.002	−0.034
25	1.212	0.051	0.004	−0.019
26	1.212	0.052	0.003	−0.026
27	1.234	0.053	0.002	−0.020
28	1.219	0.054	0.001	−0.027
29	1.226	0.053	0.002	−0.011
30	1.220	0.053	0.002	−0.031
31	1.227	0.057	−0.002	−0.018
32	1.211	0.057	−0.002	−0.018
33	1.231	0.057	−0.002	−0.029
34	1.218	0.057	−0.002	−0.027
35	1.218	0.051	0.004	−0.019

, Table 3 and Table 4 show the product compositions and densities resulting from the production of P₁, P₂, and P₃, as well as the differences between the experimental values and the theoretical values specified by the supplier.

Where D_1R represents the experimental density of P₁, Y_11R is the measured composition of the main ingredient in P₁, Y_11T − Y_11R denotes the difference between the theoretical and experimental composition values, and D_1T − D_1R indicates the difference between the theoretical and measured densities for product P₁.

Table 3. Data collection for P₂; theoretical values: D_2T = 1.330 kg/L; Y_21T = 0.600 kg C_21T/kg T; Y_22T = 0.030 kg C_22T/kg T.

Test	D2R kg/L	Y21R kg C21R/kg T	Y22R kg C22R/kg T	Y21T − Y21R kg C21/kg T	Y22T − Y22R kg C22/kg T	D2T − D2R kg/L
1	1.328	0.610	0.031	−0.010	−0.001	0.002
2	1.328	0.610	0.031	−0.010	−0.001	0.002
3	1.332	0.603	0.032	−0.003	−0.002	−0.002
4	1.309	0.596	0.031	0.004	−0.001	0.021
5	1.328	0.610	0.033	−0.010	−0.003	0.002
6	1.324	0.599	0.030	0.001	0.000	0.006
7	1.325	0.600	0.030	0.000	0.000	0.005
8	1.325	0.600	0.030	0.000	0.000	0.005
9	1.328	0.600	0.030	0.000	0.000	0.002
10	1.328	0.598	0.028	0.002	0.002	0.002
11	1.328	0.600	0.030	0.000	0.000	0.002
12	1.330	0.598	0.030	0.002	0.000	0.000
13	1.331	0.591	0.030	0.009	0.000	−0.001
14	1.333	0.602	0.031	−0.002	−0.001	−0.003
15	1.331	0.582	0.030	0.018	0.000	−0.001
16	1.332	0.592	0.032	0.008	−0.002	−0.002
17	1.335	0.598	0.031	0.002	−0.001	−0.005
18	1.330	0.593	0.028	0.007	0.002	0.000
19	1.327	0.600	0.030	0.000	0.000	0.003
20	1.336	0.600	0.029	0.000	0.001	−0.006
21	1.320	0.595	0.025	0.005	0.005	0.010
22	1.320	0.595	0.025	0.005	0.005	0.010
23	1.328	0.598	0.029	0.002	0.001	0.002
24	1.327	0.600	0.028	0.000	0.002	0.003
25	1.328	0.598	0.028	0.002	0.002	0.002
26	1.328	0.598	0.028	0.002	0.002	0.002
27	1.327	0.598	0.030	0.002	0.000	0.003
28	1.328	0.600	0.035	0.000	−0.005	0.002
29	1.330	0.600	0.030	0.000	0.000	0.000
30	1.332	0.595	0.030	0.005	0.000	−0.002
31	1.300	0.590	0.027	0.010	0.003	0.030
32	1.332	0.597	0.028	0.003	0.002	−0.002
33	1.332	0.596	0.028	0.004	0.002	−0.002
34	1.330	0.600	0.030	0.000	0.000	0.000
35	1.331	0.600	0.030	0.000	0.000	−0.001

Table 3. Data collection for P₂; theoretical values: D_2T = 1.330 kg/L; Y_21T = 0.600 kg C_21T/kg T; Y_22T = 0.030 kg C_22T/kg T.

Test	D2R kg/L	Y21R kg C21R/kg T	Y22R kg C22R/kg T	Y21T − Y21R kg C21/kg T	Y22T − Y22R kg C22/kg T	D2T − D2R kg/L
1	1.328	0.610	0.031	−0.010	−0.001	0.002
2	1.328	0.610	0.031	−0.010	−0.001	0.002
3	1.332	0.603	0.032	−0.003	−0.002	−0.002
4	1.309	0.596	0.031	0.004	−0.001	0.021
5	1.328	0.610	0.033	−0.010	−0.003	0.002
6	1.324	0.599	0.030	0.001	0.000	0.006
7	1.325	0.600	0.030	0.000	0.000	0.005
8	1.325	0.600	0.030	0.000	0.000	0.005
9	1.328	0.600	0.030	0.000	0.000	0.002
10	1.328	0.598	0.028	0.002	0.002	0.002
11	1.328	0.600	0.030	0.000	0.000	0.002
12	1.330	0.598	0.030	0.002	0.000	0.000
13	1.331	0.591	0.030	0.009	0.000	−0.001
14	1.333	0.602	0.031	−0.002	−0.001	−0.003
15	1.331	0.582	0.030	0.018	0.000	−0.001
16	1.332	0.592	0.032	0.008	−0.002	−0.002
17	1.335	0.598	0.031	0.002	−0.001	−0.005
18	1.330	0.593	0.028	0.007	0.002	0.000
19	1.327	0.600	0.030	0.000	0.000	0.003
20	1.336	0.600	0.029	0.000	0.001	−0.006
21	1.320	0.595	0.025	0.005	0.005	0.010
22	1.320	0.595	0.025	0.005	0.005	0.010
23	1.328	0.598	0.029	0.002	0.001	0.002
24	1.327	0.600	0.028	0.000	0.002	0.003
25	1.328	0.598	0.028	0.002	0.002	0.002
26	1.328	0.598	0.028	0.002	0.002	0.002
27	1.327	0.598	0.030	0.002	0.000	0.003
28	1.328	0.600	0.035	0.000	−0.005	0.002
29	1.330	0.600	0.030	0.000	0.000	0.000
30	1.332	0.595	0.030	0.005	0.000	−0.002
31	1.300	0.590	0.027	0.010	0.003	0.030
32	1.332	0.597	0.028	0.003	0.002	−0.002
33	1.332	0.596	0.028	0.004	0.002	−0.002
34	1.330	0.600	0.030	0.000	0.000	0.000
35	1.331	0.600	0.030	0.000	0.000	−0.001

Where D_2R is the experimental density of P₂; Y_21R and Y_22R represent the measured composition of the main ingredients in P₂ (two key ingredients); Y_21T − Y_21R and Y_22T − Y_22R are the differences between the theoretical and experimental composition values, and D_2T − D_2R indicates the difference between the theoretical and experimental density values for product P₂.

Table 4. Product P₃. D_3T = 1.150 kg/L; Y_31T = 0.041 kg C_31T/kg T.

Test	D3R kg/L	Y31R kg C31R/kg T	Y31T − Y31R kg C31/kg T	D3T − D3R kg/L
1	1.162	0.052	−0.011	−0.012
2	1.162	0.052	−0.011	−0.012
3	1.163	0.051	−0.010	−0.013
4	1.163	0.047	−0.006	−0.013
5	1.145	0.047	−0.006	0.005
6	1.145	0.045	−0.004	0.005
7	1.159	0.047	−0.006	−0.009
8	1.160	0.045	−0.004	−0.010
9	1.160	0.047	−0.006	−0.010
10	1.160	0.042	−0.001	−0.010
11	1.160	0.042	−0.001	−0.010
12	1.145	0.039	0.002	0.005
13	1.145	0.038	0.003	0.005
14	1.162	0.041	0.000	−0.012
15	1.142	0.044	−0.003	0.008
16	1.142	0.044	−0.003	0.008
17	1.144	0.045	−0.004	0.006
18	1.144	0.045	−0.004	0.006
19	1.149	0.047	−0.006	0.001
20	1.163	0.041	0.000	−0.013
21	1.161	0.047	−0.006	−0.011
22	1.161	0.049	−0.008	−0.011
23	1.160	0.045	−0.004	−0.010
24	1.145	0.045	−0.004	0.005
25	1.145	0.045	−0.004	0.005
26	1.153	0.047	−0.006	−0.003
27	1.160	0.048	−0.007	−0.010
28	1.160	0.048	−0.007	−0.010
29	1.172	0.048	−0.007	−0.022
30	1.173	0.050	−0.009	−0.023
31	1.173	0.052	−0.011	−0.023
32	1.179	0.049	−0.008	−0.029
33	1.157	0.046	−0.005	−0.007
34	1.172	0.040	0.001	−0.022
35	1.153	0.037	0.004	−0.003

Table 4. Product P₃. D_3T = 1.150 kg/L; Y_31T = 0.041 kg C_31T/kg T.

Test	D3R kg/L	Y31R kg C31R/kg T	Y31T − Y31R kg C31/kg T	D3T − D3R kg/L
1	1.162	0.052	−0.011	−0.012
2	1.162	0.052	−0.011	−0.012
3	1.163	0.051	−0.010	−0.013
4	1.163	0.047	−0.006	−0.013
5	1.145	0.047	−0.006	0.005
6	1.145	0.045	−0.004	0.005
7	1.159	0.047	−0.006	−0.009
8	1.160	0.045	−0.004	−0.010
9	1.160	0.047	−0.006	−0.010
10	1.160	0.042	−0.001	−0.010
11	1.160	0.042	−0.001	−0.010
12	1.145	0.039	0.002	0.005
13	1.145	0.038	0.003	0.005
14	1.162	0.041	0.000	−0.012
15	1.142	0.044	−0.003	0.008
16	1.142	0.044	−0.003	0.008
17	1.144	0.045	−0.004	0.006
18	1.144	0.045	−0.004	0.006
19	1.149	0.047	−0.006	0.001
20	1.163	0.041	0.000	−0.013
21	1.161	0.047	−0.006	−0.011
22	1.161	0.049	−0.008	−0.011
23	1.160	0.045	−0.004	−0.010
24	1.145	0.045	−0.004	0.005
25	1.145	0.045	−0.004	0.005
26	1.153	0.047	−0.006	−0.003
27	1.160	0.048	−0.007	−0.010
28	1.160	0.048	−0.007	−0.010
29	1.172	0.048	−0.007	−0.022
30	1.173	0.050	−0.009	−0.023
31	1.173	0.052	−0.011	−0.023
32	1.179	0.049	−0.008	−0.029
33	1.157	0.046	−0.005	−0.007
34	1.172	0.040	0.001	−0.022
35	1.153	0.037	0.004	−0.003

In the same way, D_3R is the experimental density of P₃, Y_31R is the measured composition of the main ingredient in P₃, Y_31T − Y_31R indicate the difference between the theoretical and experimental values of compositions, and D_3T − D_3R represent the difference between the theoretical and experimental density values for product P₃.

Table 5. Average values (x), lower limits (LI), upper limits (LS), percentage variations with theoretical values (Δ), and standard deviations (σ) of parameters analyzed for P_1, P₂, and P₃.

Pi	x	Δx (%)	LI	ΔLI (%)	LS	ΔLS (%)	σ
D1R	1.222	1.83	1.208	0.67	1.234	2.83	0.007
Y11R	0.054	−1.82	0.047	−14.55	0.057	3.64	0.002
D2R	1.327	−0.23	1.300	−2.26	1.336	0.45	0.007
Y21R	0.598	−0.33	0.582	−3.00	0.610	1.67	0.005
Y22R	0.030	0.00	0.025	−16.67	0.035	16.67	0.002
D3R	1.157	0.61	1.142	−0.70	1.179	2.52	0.010
Y31R	0.046	12.20	0.037	−9.75	0.052	26.83	0.004

shows the mean values, upper and lower limits, and standard deviations for each data series. These metrics can provide a concise summary of the dataset, offering a clear understanding of the tendency, variability, and value range.

Where x is the average value, Δx represents the percentage variation from the theoretical values, LI and LS are the lower and upper limits of the interval, ΔLI and ΔLS represent the percentage variation from the theoretical values, and σ is the standard deviation of each variable.

3.3. Analyze

Table 6. Total product loss (ΔP_T) and raw material loss (ΔMP_ij).

Pi	PT L	$\overline{D_{iT}-D_{iR}}$ kg/L	ΔPT kg	$\overline{Y_{ijR}}$ kg CijR/kg T	ΔYij kg CijR/kg T	ΔMPij kg
P1	193,440	−0.022	−4255.68	0.054	−229.81	−919.23
P2	72,160	0.003	216.48	0.598 0.030	129.46 6.49	199.16 25.98
P3	130,080	−0.007	−910.56	0.046	−41.89	−598.43

presents the estimated material loss for each product resulting from composition and density variations during the manufacturing process.

Where P_T is the total volume produced in the 35 batches of each product, $\overline{\left({\mathrm{D}}_{\mathrm{iT}}-{\mathrm{D}}_{\mathrm{iR}}\right)}$ represents the average difference between the theoretical and experimental density values, $\overline{{\mathrm{Y}}_{\mathrm{iJR}}}$ is the average composition, and ΔY_ij indicates the loss of each ingredient. These data were used to calculate the total raw material loss (ΔMP_ij) and the total mass loss (ΔP_T). Thus, the total amount of raw material lost across the 35 batches was 919.23 kg for P₁, 199.16 and 25.98 kg for P₂, and 598.43 kg for P₃. It should be noted that the total quantity of the products manufactured in the 35 batches accounts for 23% of the company’s global production during the 12-month measuring period. Furthermore, the financial loss incurred by these productions was estimated to be EUR 35,500.00, representing 5% of the total sales revenue generated by the products in question during the specified period.

3.4. Improve

The results indicate that the most significant deviations occurred in those products where the analyzed component was supplied in a solid form and in large quantities of liquid, i.e., P₁ and P₃, respectively (P₂ showed negative deviations, Table 5 and Table 6). Therefore, the improvement activities were focused on products P₁ and P₃. The main objective was to reduce the variability in the composition of P₁ and P₃ by 50%. This reduction can be achieved by minimizing the variability in product density, which is directly correlated with the loss of raw materials.

Table 7. Target values for a 50% reduction in raw material and product loss.

Loss estimation P1, kg	−2127.84
Average value (D1T − D1R), kg/L	−0.011
Estimated average value D1R, kg/L	1.211
Variation of D1R over D1T, %	0.92
Loss estimation MP11, kg	−468.97
Loss estimation Y11, kg C11/kg tot	−114.91
Loss estimation P3, kg	−459.62
Average value (D3T − D3R), kg/L	−0.004
Estimated average value D3R, kg/L	1.154
Variation of D3R over D3T, %	0.35
Loss estimation MP31, kg	−299.22
Loss estimation Y31, kg C31/kg tot	−20.95

displays the target values required to achieve a 50% reduction in product variability.

Consequently, the density target values for the P₁ and P₃ were 1.211 and 1.154 kg/L, respectively. Manufacturing products with these densities will result in a 50% reduction in the composition variability of both products.

3.5. Control

According to the value stream map analysis, the material handling stages were identified as the key areas for optimization. Additionally, the Ishikawa diagram (Figure 2) outlined the critical factors affecting product variability, including methods, environment, facilities, materials, staff, and measurements. Therefore, a series of improvement initiatives were designed to optimize these points.

Table 8. Actions taken following the Ishikawa diagram analysis.

Category	Causes	Action	Involved	Result/Cost	Time
Methods	Production programming	MP production and procurement plan Demand for updated sales forecasts	Planning and Purchasing Commercial	Gant diagrams. ERP adaptationCRM EUR 5000	3–6 months
	Standardization manufacturing procedures	Specific working instructions	Production	Detailed working procedure	2 months
Facilities	Equipment availability	Specific preventive maintenance plan	Maintenance	Technical services MES 4500 €	3–6 months
		Timetable with prioritization of productions	Production	Production program	-
Staff	Appropriate staff training	Training on product handling and time management	Human Resources	Training program EUR 3000	1 week
	Precise raw materials handling	Training in handling measuring equipment	Production and Human Resources	Training program EUR 1500	1 week
Environment	Workspace layout	Elimination of obstacles Tidiness of the work area	Production Production	Reorganization of spaces Good practices	1 week
	Delimit a suitable and specific area for raw materials	Raw materials location área Regular audits	Maintenance	Signaling and use of beacons EUR 200	2 weeks
Materials	Reduce variability of raw materials composition	Choice of raw materials whose variability in composition is <2%	Purchasing	Search for three suppliers and choose one with the least variability in composition	1–2 months
Measurement	Verification of measurement equipment	Establish a program for the adjustment and calibration of measuring equipment	Quality	Entity official verification EUR 500	1 month

shows the improvement activities carried out in the company to reduce product variability. Briefly, the activities involved the establishment of a procurement plan and sales forecasts to ensure a steady supply of raw materials, aligning procurement with production needs and market demand. Specific working instructions were also developed to standardize operator tasks, minimizing human error and maintaining consistency. Moreover, a maintenance plan was put in place to ensure equipment reliability and prevent any disruptions that could affect product quality. Additionally, production scheduling was optimized to ensure smooth workflows and reduce bottlenecks. The company also introduced training plans to enhance employees’ skills and ensure accurate task execution, reducing errors.

To illustrate an example, Figure 5 presents a scheme of the initial warehouse disposition and the changes adopted to improve the location and management of raw materials and the operating area.

Upon arrival, the raw materials were stored in a designated area for quality control purposes. However, due to limited space and disorganized worksite logistics, materials were often stored in a non-ordered manner, leading to confusion and wasted time (Figure 5a). As seen in Figure 5b, a specific area was designated for the proper classification of materials based on their intended use. This approach also facilitated the placement of pre-weighted materials in storage locations adjacent to the manufacturing area, reducing the time required for material preparation by 42%. Furthermore, clean warehouse practices were introduced, along with bi-monthly audits, to ensure ongoing organization and efficiency (Table 8).

In order to assess the effectiveness of the proposed activities and the acceptability of the implemented solutions, a total of 15 runs for each product were carried out. The total production in these batches was 300,000 L (10,000 L/batch), which is equivalent to the production from the 35 batches used for the initial measurements.

Table 9. Experimental values derived from the production of P₁.

Test	D1R’ kg/L	D1T − D1R’ kg/L	Y11R’ kg C11/kg T
1	1.200	0.000	0.049
2	1.215	−0.015	0.053
3	1.220	−0.020	0.054
4	1.199	0.001	0.051
5	1.196	0.004	0.050
6	1.215	−0.015	0.052
7	1.225	−0.025	0.053
8	1.210	−0.010	0.051
9	1.220	−0.020	0.053
10	1.215	−0.015	0.053
11	1.195	0.005	0.050
12	1.198	0.002	0.051
13	1.200	0.000	0.050
14	1.212	−0.012	0.052
15	1.210	−0.010	0.052

and Table 10 show the densities (D_1R’, D_3R’) and compositions (Y_11R’, Y_31R’) obtained from the production of P₁ and P₃ after the application of LSS tools.

Where D_1R’ is the experimental density, Y_11R’ is the measured composition of the main ingredient in P₁, and D_1T − D_1R’ is the difference between the theoretical and experimental density values for product P₁.

Table 10. Experimental values derived from the production of P₃.

Test	D3R’ kg/L	D3T − D3R’ kg/L	Y31R’ kg C31/kg T
1	1.140	0.010	0.043
2	1.165	−0.015	0.045
3	1.150	0.000	0.044
4	1.155	−0.005	0.047
5	1.175	−0.025	0.049
6	1.160	−0.010	0.047
7	1.166	−0.016	0.046
8	1.170	−0.020	0.048
9	1.152	−0.002	0.044
10	1.160	−0.010	0.045
11	1.145	0.005	0.043
12	1.150	0.000	0.044
13	1.155	−0.005	0.045
14	1.161	−0.011	0.046
15	1.160	−0.010	0.045

Table 10. Experimental values derived from the production of P₃.

Test	D3R’ kg/L	D3T − D3R’ kg/L	Y31R’ kg C31/kg T
1	1.140	0.010	0.043
2	1.165	−0.015	0.045
3	1.150	0.000	0.044
4	1.155	−0.005	0.047
5	1.175	−0.025	0.049
6	1.160	−0.010	0.047
7	1.166	−0.016	0.046
8	1.170	−0.020	0.048
9	1.152	−0.002	0.044
10	1.160	−0.010	0.045
11	1.145	0.005	0.043
12	1.150	0.000	0.044
13	1.155	−0.005	0.045
14	1.161	−0.011	0.046
15	1.160	−0.010	0.045

Similarly, D_3R’ indicates the experimental density, Y_31R’ is the measured composition of the main ingredient in P_3, and D_3T − D_3R’ is the difference between the theoretical and experimental density values for product P₃.

For a better interpretation of the results, the target density and composition values to achieve a reduction of 50% in product variability were estimated for each product and subsequently compared with the experimental values. The estimated density values (D_1R* and D_3R*) were determined using the average density values from Table 7 (0.011 and 0.004) and the measured density values (D_1R and D_3R) from Table 2 and Table 4, respectively. Similarly, the desired composition values were determined (Y_11R* and Y_31R*) for each product.

Figure 6 represents the data obtained after the application of LSS tools, i.e., D_1R’, Y_11R’ (Table 9) and D_3R’, Y_31R’ (Table 10), and the estimated values required to achieve a 50% reduction in product variability, i.e., D_1R*, Y_11R*, D_3R*, and Y_31R* (

Table 11. D_1R*, D_3R*, Y_11R*, and Y_31R* estimated values for 50% loss reduction.

P1		P3
D1R* kg/L	Y11R* kg C11/kg T	D3R* kg/L	Y31R* kg C31/kg T
1.211	0.054	1.158	0.057
1.211	0.055	1.158	0.057
1.215	0.054	1.159	0.056
1.218	0.056	1.159	0.052
1.216	0.056	1.141	0.052
1.208	0.050	1.141	0.050
1.197	0.054	1.155	0.052
1.202	0.052	1.156	0.050
1.218	0.053	1.156	0.052
1.199	0.046	1.156	0.047
1.215	0.051	1.156	0.047
1.215	0.050	1.141	0.044
1.209	0.050	1.141	0.043
1.216	0.056	1.158	0.046
1.219	0.055	1.138	0.049
1.217	0.056	1.138	0.049
1.214	0.051	1.140	0.050
1.201	0.051	1.140	0.050
1.203	0.051	1.145	0.052
1.219	0.054	1.159	0.046
1.221	0.054	1.157	0.052
1.206	0.051	1.157	0.054
1.215	0.050	1.156	0.050
1.222	0.050	1.141	0.050
1.201	0.053	1.141	0.050
1.201	0.051	1.149	0.052
1.223	0.056	1.156	0.053
1.208	0.050	1.156	0.053
1.215	0.051	1.168	0.053
1.209	0.052	1.169	0.055
1.216	0.053	1.169	0.057
1.200	0.052	1.175	0.054
1.220	0.052	1.153	0.051
1.207	0.056	1.168	0.045
1.207	0.056	1.149	0.042
1.211	0.054	1.158	0.057

).

As can be seen, the values obtained after implementing the LSS framework fall within the range of the estimated values required to achieve a 50% reduction in variability. This proves the effectiveness of the improvement actions proposed, successfully fulfilling the research objective.

Table 12. Upper limits (UL), lower limits (LL), average values (x), and standard deviations (σ) for each parameter.

	D1R*	Y11R*	D1R’	Y11R’	D3R*	Y31R*	D3R’	Y31R’
UL	1.223	0.056	1.225	0.054	1.175	0.057	1.175	0.049
LL	1.197	0.046	1.195	0.049	1.138	0.042	1.140	0.043
x	1.211	0.053	1.209	0.052	1.153	0.051	1.158	0.045
σ	0.007	0.002	0.010	0.001	0.010	0.004	0.009	0.002

shows the upper limits (ULs), lower limits (LLs), average values (x), and standard deviations (σ) for each parameter obtained using the data from Table 9, Table 10 and Table 11.

The experimental composition values for P₁ and P₃ (Y_11R’, Y_31R’) exhibited a significant decrease in variability, with reductions of 50% and 40%, respectively, compared to the estimated values (Y_11R*, Y_31R*). Additionally, the average value (x) for P₁ and P₃ decreased by 1.89% and 11.77%, respectively. These results led to a 38% decrease in global material overuse costs. The standard deviations and mean values of the data define the arguments of the normal distributions, enabling the determination of the probability of each data occurrence, as expressed by Equation (12).

$${\phi }_{\overline{x},\sigma }\left(x\right)={\displaystyle \frac{1}{\sigma \sqrt{2\pi }}}{e}^{-{\displaystyle \frac{(x-{\overline{x)}}^2}{2{\sigma }^2}}}$$

(11)

where ${\phi }_{\overline{x},\sigma }\left(x\right)$ represents the density function. The graphical representation of this normal distribution can be expressed in terms of nσ, where −6 < n < 6 covers the behavior of the data within the Six Sigma range and is also standardized by dividing by the maximum value of the density function obtained (N(x, nσ)). Figure 7 presents the normal distribution plots for the experimental values of P₁ (I, II) and P₃ (III, IV) products before and after LSS application, along with the target values to achieve a 50% reduction in variability.

It is evident that the experimental data obtained after the application of the LSS tools (dotted line) closely align with the target values necessary to achieve the desired reduction in variability (orange line). Additionally, the experimental data obtained after the application of LSS tools (dotted line) show a narrower amplitude of the typical Gaussian curve than those obtained prior to the implementation of LSS tools (blue line). In the case of Y₁ and Y₃ values (Figure 7II,IV), the experimental data obtained after implementing LSS tools show a distribution nearly identical to the target data required to achieve a 50% reduction in product variability; however, there are slight variations in the values of D₁ and D₃ (Figure 7I,III). Thus, it can be stated that the improvement actions had a greater influence on composition (Figure 7I,III) than on density (Figure 7II,IV). This reduction in product variability resulted in a 39.7% decrease in costs from product loss and a 53.7% decrease in costs from raw material loss. Therefore, it can be concluded that the standardization of manufacturing procedures becomes a milestone in achieving uniformity in the production process.

Similar results have been reported by several authors applying similar improvement actions in different industries. For example, Meeuwse [35] conducted a case study on a chemical company that enhanced a multistep batch process by analyzing the duration of relevant process stages and optimizing procedures. The plant productivity increased by over 20% through procedural changes and reprogramming of the process control system. Silva et al. [36] applied project management tools to improve planning and time control to mitigate delivery delays in the metalworking industry and reported a 40% reduction in the average time deviations. Lu and Liu [37] improved the operation of a raw material warehouse in a logistics company. They achieved a reduction in non-value-added time by up to 36.7% and a reduction in supply cycle time by up to 21%, along with a 5.7% increase in process cycle efficiency. In a case study conducted by Realyvásquez Vargas [38], a manufacturing company improved its raw materials receiving process in the warehouse area, achieving an 85% reduction in processing time. Additionally, the capacity for labeling and identification operations increased from 580 to 1470 units (253.4% increase), while the number of process activities was reduced from 37 to 20 (45.9% reduction).

4. Conclusions

This work demonstrates the successful application of LSS tools in an industry and process that has remained almost unexplored until now. An LSS framework was developed to identify and reduce sources of variability occurring in the final product composition of a Spanish SME fertilizer manufacturing. The main variables affecting the product variability, i.e., density and ingredient composition, were determined using the Ishikawa, SIPOC diagrams, and VSM. The LSS framework was implemented according to the DMAIC steps, and the results were validated using statistical tools. The application of LSS tools enhanced the overall operational efficiency by reducing product variability by 50% and material preparation time by 42%. This reduction led to a decrease in costs from product losses of 40% and a decrease in costs associated with raw material losses of 54%. The LSS framework applied in this study is limited to controlled batch productions, which may not fully represent the variability and challenges of other complex chemical processes. Furthermore, the case study is based on a single company, which limits the generalizability of the findings. Despite these limitations, this study offers valuable insights and can serve as a reference for future research, particularly given the limited literature available on its application within the chemical manufacturing industry. Thus, there is a significant potential for future research to expand the understanding of LSS adoption across diverse chemical manufacturing processes, as well as in other specific sectors.