This research aims to enhance digital marketing processes by using design of experiments in safelist mailers and traffic exchange websites.
The research applies design of experiments to improve an online digital marketing website.
The findings show that design of experiments can help improve digital marketing quality by generating more interactions with safelist mailers and traffic exchange websites, reducing time spent in the marketing process, allocating the appropriate amount of credits to significant websites and avoiding product price increases.
This research aims to increase awareness of the impact and applicability of design of experiments to digital marketing and demonstrates the application of design of experiments in any digital marketing process, including safelist mailers and traffic exchange websites.
To the best of the authors’ knowledge, this is the first study applying design of experiments within online marketing. This study can be leveraged by academics and marketing functions to demonstrate the benefits of design of experiments to the marketing function to improve process efficiency and resource utilisation.
1. Introduction
The traditional purpose of marketing has been to achieve success in sales, market share and gross margin in the marketplace (Weber and De Villebonne, 2002). Therefore, organisations must understand the benefits of efficient marketing processes to keep or gain more market share and have stronger customer relations. However, unfortunately, companies have not yet successfully created practical new approaches for measuring and improving marketing productivity (Oliya et al., 2012), and it is not always easy to measure this (Yadava et al., 2022).
With recent online retail advances, online marketing has become very popular (Angeloni and Rossi, 2021), and since the global COVID-19 pandemic, consumers are increasingly shopping online in preference to the conventional method of physical shopping. Thus more retailers are replacing online shops and retail outlets with online stores requiring marketing strategies to attract and retain customers (Grewal et al., 2003). While organisations now regard having a website as part of their normal business, the real challenge is attracting traffic to a specific website (Quinton and Khan, 2009). As marketing processes can have many steps and inputs process optimisation naturally lends itself and complements marketing processes by aiding in the analysis of inefficiencies and in capturing the voice of the customer(Mohamad and Hasan, 2024; Sá et al., 2022).
Many continuous improvement methods are utilised by organisations in all industries, with Lean, Six Sigma, and Lean Six Sigma (LSS) evolving as the continuous improvement (CI) methodologies of choice in recent years (Byrne et al., 2021). Lean, focusing on reducing waste and Six Sigma, focusing on variation, are utilised together to enhance processes and improve customer satisfaction (George, 2002).
LSS has been deployed in manufacturing functions extensively (McDermott et al., 2022; Snee and Hoerl, 2018) and, for example, within the design and new product functions (Slattery et al., 2022), human resources (Trubetskaya and Mullers, 2021) and procurement functions (Suominen, 2020). However, while CI methodologies have been deployed widely in many organisations and functions, the application in marketing is not as researched or studied as in other sectors (Chaplin and O’Rourke, 2014; Payaro and Papa, 2016; Sá et al., 2022). This lack of research is primarily due to the predominance of LSS application in manufacturing functions and operations as Lean and Six Sigma is derived from the manufacturing floor (Snee and Hoerl, 2018).
Design of experiments (DOE) as an important statistical tool in the LSS toolkit aids in the planning, conducting and analysing of experiments systematically and statistically to improve products and processes. However despite its value, Lundkvist et al. (2020) highlight that DOE is the least-utilised statistical method as opposed to statistical process control (SPC) and process capability analysis. They further stated that insufficient resources (e.g. time and money), lack of management commitment and poor statistical knowledge to aid application in experimentation are the reasons for the lack of utilisation of DOE.
Pestorius (2007) also provides reasons as to why DOE application to Sales and Marketing are uncommon; (1) It is believed that almost every variable can be controlled in manufacturing processes. (2) Process improvement consultants are typically from manufacturing and do not understand transactional processes such as sales and marketing. (3) There is not as much pressure in relation to quality concerns and global pressures as there is in manufacturing. (4) A marketing culture of individual and entrepreneurial thinking actively resists standardised processes. It is only in recent years that authors have researched and carried out case study applications of statistical tools in marketing functions. As marketing is generally closest to the customer, they are a natural stakeholder in continuous improvement projects to represent the Voice of the Customer (VOC) and create value for customers (Chaplin and O’Rourke, 2014; Sanchez, 1999). Process improvement in sales and marketing is important for the overall performance and growth of business as it leads to improved market performance, competitiveness and superior value creation (Madhani, 2017). However, studies in the literature tend to refer to the absence of marketing as a stakeholder in continuous improvement projects and their lack of embracing of continuous improvement tools rather than putting forward case studies on continuous improvement applications in Marketing processes and activities (Chaplin and O’Rourke, 2014; Quelch and Harris, 2005; Strategic Direction, 2014). Indeed, Chaplin and O’Rourke described using continuous improvement tools in in marketing as a “missed opportunity”.
The case study organisation where this research takes place utilised online traffic exchanges (TE) and safelist mailers (SM). TE services enable members to bring traffic to their websites from a diverse pool of IP addresses in return for visiting the sites of other members (Javed et al., 2015). It is not always easy to verify if computer users see these websites when working in the background. In digital marketing, hits are needed, but visits are important to increase sales (Kalyanam and McIntyre, 2002). The problem statement is with the use of TE and SM is that a lack of utilisation, non-value-add steps and poor processes in the set-up of SM and TE websites means fewer clicks and interactions with products from a marketing point of view, thus decreasing revenue opportunities. The utilisation and application of LSS techniques can identify the best methods of optimising the TE and SM set up processes and increase revenue opportunities. DOE lends itself to aiding with process optimisation and improving efficiency (Antony et al., 2024).
Thus, the research question (RQ) is to ascertain if DOE can be utilised to help improve the quality of the digital marketing process. In the case study website for this research, an online store named Julsine is trying to increase revenue and reduce waste and insufficiencies around SM’s and TE’s website marketing processes and usage. According to Six Sigma Daily (2019), the few works applying continuous improvement tools in digital marketing deal with search engines, emails, social media and video. This study will demonstrate a new paradigm in DOE and its application to Marketing. No studies thus far have dealt with applying DOE to safelist mailer and traffic exchange websites; thus, this study will be a pilot within the literature. Generally, DOE application has been found to be sparse in all types of industries and sectors (Antony et al., 2024; Frey and Haller, 2021), so a case study demonstration of its effectiveness can aid increased application and knowledge of DOE.
The project will generate more interactions with safelist mailers and traffic exchanges websites, thus minimising time spent in the marketing process and maximising the use of the credits needed for promotion. In addition, the research will further the knowledge of the digital marketing area and how DOE can be applied to aid digital marketing (Kankam and Dza, 2023). Design of experiments is a well-established protocol for designing, developing and improving products and processes, which can be used to help achieve marketing’s goals of delighting the customer and achieving market share. Thus, the study will fill a gap in research by helping academicians, practitioners and marketers to benchmark a comprehensive, structured and multi-disciplinary approach to waste reduction and optimisation of resource utilisation in the marketing space.
The remainder of the paper is as follows; Section 2 describes the literature and the research methodology in Section 3. Next, the results are presented in Section 4, followed by a discussion and implications in Section 5. Finally, the conclusion, limitations and scope for future research are outlined in Section 6.
2. Literature review
According to Clifford (2001) and Bala and Verma (2018) while traditional marketing channels remain a big part of marketing campaigns, the real action is in digital marketing, leveraging technology that did not exist traditionally to reach audiences in new ways (Wertime and Fenwick, 2011). Digital marketing encompasses all marketing effects that use an electronic device or the internet.
The term “digital marketing” has “evolved over time from describing the marketing of products and services using digital channels to a term describing the process of using digital technologies to acquire customers and build customer preferences, promote brands, retain customers and increase sales” (Kannan and Li, 2017).
Moreover, businesses leverage digital channels such as safelist mailers, search engines, social media, traffic exchanges and other websites to connect with current and prospective consumers (Ryan, 2016). Digital marketing, also called inbound marketing, concentrates on attracting customers rather than interrupting them. As a result, consumers are empowered to research online as they progress through their buyer journey (Adam et al., 2020; Cole et al., 2017; Lincoln, 2019).
Businesses are using various channels: affiliate marketing, email marketing, search engine marketing (SEM), video marketing, safelist mailers and traffic exchange websites to maximise consumer sales and profit by distributing goods to the ultimate consumers at minimum cost (Smith, 2011).
2.1 Safelist mailers and traffic exchanges websites
An internet marketing safelist is a list of email addresses that belong to people who have agreed to accept mailings. Safelist members can often find their inboxes bombarded with hundreds of daily emails (Duffy, 2007). Sometimes these are from marketers trying to build their lists by offering downloads and other freebies, but more often than not, the mailings will be from internet marketers hoping to make a sale.
Traffic exchanges have been categorised as having a manual or auto-surf exchange. The user manually clicks through the other member’s websites or ads in the manual surf exchange. The auto-surf exchange is easier to use but far less effective. In essence, it runs automatically without effort on the user’s part, changing websites or ads every few seconds. If the user does not even need to look at it to have it work, people will likely not view their ads (Marius, 2014).
Many authors, including Jackson (2022), and Jarrett (2007), maintain that many safelists and traffic exchange websites are free to join, but some of the supposedly better ones may require new users to pay a one-off membership fee (Duffy, 2007). With others, continued membership may require a monthly payment, and some are more effective and relevant than others (Michie, 2006).
In simple terms, an internet marketing safelist is a list of email addresses that belong to people who have agreed to accept mailings (Marius, 2014). While with traffic exchanges websites, the basic idea is to view other members’ websites or ads (Javed et al., 2015). Unfortunately, the rotation of some of these websites does not generate enough hits and visits to the store site. The internet is full of websites that use automated, black hat, or bot software to generate clicks (Di Fatta et al., 2018).
Moreover, the processes that use these websites waste valuable time, and the cost of credits needed for the promotion unnecessarily increases the price of the products and services (Clifford, 2020). The consequence is that the statistics show only the hits without effective visits to the site (Novak and Hoffman, 1997). As a set of tools and strategies to improve processes outputs, Lean Six Sigma might provide the framework needed to succeed (Chaplin and O’Rourke, 2014).
The application of Lean Six Sigma to digital marketing is relatively new (Kankam and Dza, 2023). As a result, digital marketing retailers face the challenge of designing an environment that can engage prospects and customers within a few minutes of a single navigation with different platforms running simultaneously in different areas (Demangeot and Broderick, 2016).
This requires both proper knowledge and management to represent the online marketing environment in a better way (Calder et al., 2009). Moreover, the users of almost every company’s products and services are shifting in makeup, location and number at an ever-increasing rate. Thus, marketers must revise yearly strategies to keep up with the latest best practices. Many of the processes created in digital marketing are unrefined.
2.3 Lean six sigma
Six Sigma deployment can emphasise performance measurement to sales and marketing activities and add measurable value to sales and marketing performance to help increase market share and top-line revenue in targeted products/markets (Madhani, 2017). However, few organisations have embraced Six Sigma in their marketing functions, but that is changing. For example, Xerox has applied Six Sigma in its marketing team, experienced its effectiveness and had 80 Six Sigma Black belts in its sales and marketing group (Calabro, 2004). In addition, Bank of America and National City Corporation, 3M, Allied Signal, Dell, HSBC, Johnson and Johnson, LSI Logic, Owens Corning, Service Master and Standard Register (Taylor Communications) have embraced the application of Six Sigma to operations, and IT functions for several years and are now recognising its place in marketing (Madhani, 2017).
The contribution of Lean combined with Six Sigma consists of continual improvement based on increasing customer value. Marketeers being uniquely focused on customer value above all other functionals naturally aligns with LSS. Sá et al. (2022) have utilised LSS to improve marketing processes in a metalworking company, while Muralidharan and Raval (2020) state that sales and digital marketing’s productivity have become a vital component of Lean Six Sigma quality implementation. Based on how the services sector has introduced LSS methodologies into their management, there is evidence of the success of LSS in marketing. Furterer (2009) and Antony et al. (2017) sustain that Lean Six Sigma has broad applicability, including manufacturing and services like marketing. Starkey et al. (1997) suggest that implementing Lean Six Sigma has allowed digital marketers to learn that process improvement has advantages far beyond the manufacturing sector where it began. Thus, applying Lean Six Sigma helps digital marketing improve its marketing campaigns.
2.4 Design of experiments in marketing
In relation to traditional Six Sigma statistical analysis, there is also very little use of design of experiments or DOE in marketing. Starkey et al. (1997) in their study gave an example showing how DOE could apply to marketing by designing an effective direct-response television advertisement. As early as 1965 and 1975, there were studies applying DOE to marketing. Chevalier designed a factorial experiment to measure the impact of in-store displays on sales for different product characteristics and Hoofnagle examined the effectiveness of different facets of promotion and advertising utilising DOE (Chevalier, 1975; Hoofnagle, 1965). However, the limited works applying Lean Six Sigma in digital marketing deal mostly with search engine marketing, email marketing, social media marketing and video marketing, and there is even more sparse application of statistical six sigma tools.
The review has highlighted the need for more research to assess how DOE can continuously improve and aid the quality of digital marketing. Given the sparse studies on the application of DOE methods to the Marketing function, this study will demonstrate the practical application of DOE in an online store.
3. Methodology
This study will be a case study in an online marketing store. The background of the study and the methods deployed are described below.
3.1 Background to study
This study is a case study of an online store called Julsine. A single case study was deemed the best method to deploy and test applying DOE methods in a specific context of a single organisation. According to Yin (2011) case studies are appropriate when the themes under investigation is exploratory and when there is a dearth of knowledge on the subject. In this case there is a dearth of knowledge in the area of DOE application in online marketing. Also a single site case study in particular when the concept under research and investigation is novel and revelatory (Dubé and Paré, 2003). The organisation analysed is an owner-managed online store and thus the researchers had the advantage that were given full permission and access to data to complete the research as a single case study site. Also, a single case study allowed the involvement of the entire store to aid the study and offered a proof of concept. The single-case study allowed capture of data and sourcing of explanations for the different interdependencies and interactions within a particular context (Retolaza and San-Jose, 2017). As the researchers were given unprecedented access to the site and data as single-case researchers they were able to apply the theory and tools as per the theoretical framework to the site. This single case study access allowed the “making sense” of the empirical data and development of theory, specifically in this online store. Developing theory based on single-case research provides the researchers with unique opportunities to ground the meaning of concepts in empirical observation, experimentations and description (Anderson et al., 2020). This online store was chosen as a suitable single case study as there was an opportunity to gather information from the online store members and aid the research objective (Retolaza and San-Jose, 2017).
This study will apply DOE in particular as an improvement methodology to provide the framework needed to improve safelist mailers and traffic exchanges websites. Variability is one of the sources of waste within a process, so DOE as a Six Sigma tool can aid variability reduction while the application of Lean reduces waste (George, 2002). Moreover, considering Julsine Store’s case, a marketing process exists but does not apply the Lean Six Sigma methodology. Therefore, the researcher will apply DOE within a define, measure, analyse, improve and control (DMAIC) methodology and deploy DOE statistical tools utilising Minitab.
Additionally, the DMAIC cycle provides the opportunity to materialise the “how to go about it” to resolve marketing site optimisation with minimum resources (Besseris, 2011). Within this study utilising DOE, the key part of the DMAIC that is important is the “analyse” and “improve” stage. Indeed Boyles (2010) referred to “DOE – the power tool of the DMAIC Analyse and Improve Phases”.
Thus, DOE tools will be used to improve digital marketing within the “Julsine” Online Store. As mentioned above, several safelist mailers and online traffic exchange websites exist. The store uses many of them. Seven websites have been chosen based on the following criteria: promotion of both websites links and texts ads, the offer of low-cost credits needed for the promotion and the granting of credits when surfing other marketers’ ads.
This article aims to improve the quality of digital marketing, especially the safelist mailers and traffic exchanges websites. The critical to quality (CTQ) characteristics are essential in improving processes (Hakimi et al., 2018). This case study focuses on the CTQ to enhance digital marketing within Julsine Store. The CTQs for this study were identified as (1) generate several visits to the Store websites, (2) eliminate the time spent on unnecessary marketing efforts and (3) allow avoiding unnecessary product prices increase.
3.2 DMAIC
While this study is primarily a DOE application, the Six Sigma DMAIC methodology was followed to structure the marketing site DOE application and improvement. While DOE and analysis of variance are not new, they remain “vital ingredients” in virtually all applications of the Lean Six Sigma DMAIC cycle (Boyles, 2010).
3.2.1 Define the problem statement and measure the data.
The online store Julsine uses various digital marketing channels to promote its activities: safelist mailer (SM) and traffic exchanges (TE) websites. The use of these channels requires the following steps:
registration for membership, including membership fees payment;
add links (Sites, Texts Ads, Banners and Splash Pages) to the websites for promotion;
purchase of credits for rotation;
allocation of credits to links (Sites, Ads, Banners and Splash Pages); and
activation of links for rotation through websites.
The store is experiencing the difficulty of lack of visits from traffic exchanges and safelist mailers’ websites. This work aims to identify the combinations of these websites that generate the highest visit figures. This will eliminate inefficient digital marketing channels and reduce the time and money spent on unnecessary marketing efforts. For the seven websites used, over €2,132 was spent on purchasing 70,000 credits and 1 h was spent in the marketing process, with the site recording 150 clicks and 120 visits after 48 h. The store wants to increase the clicks and the visits while using the same number of credits and spending less than 1 h in the marketing process. As part of the measure phase, data was collected and measured on various website combinations and outcomes to detect their marketing impact quality. To achieve the study’s objectives, the researcher measured SM and TE’s websites’ impacts on the site’s hits and visits.
3.2.2 Analyse, improve and control.
The next step was to analyse trends or correlations to identify the websites that yielded the highest visits to the online website. In this stage, Minitab software and design of experiments (DOE) are used to perform the analysis. The American Society of Quality (ASQ) defines DOE as a branch of applied statistics that deals with planning, conducting, analysing and interpreting controlled tests to evaluate the factors influencing a parameter’s value or group of parameters (ASQ, 2023). Moreover, DOE is one of the most powerful quality improvement techniques for reducing process variation and enhancing process effectiveness and capability in the twenty-first century (Antony, 2014). Classical DOE emphasises the development and use of regression models to predict the process behaviour under various process conditions at a certain level of confidence (Antony, 2006). Therefore, classical DOE has been chosen as an approach to develop mathematical models connecting the responses and a set of process parameters and their interactions and determining the optimal response for the digital marketing process within the store. In the improve phase of the study, it will be determined where time and money are spent on unnecessary marketing efforts based on the analysis results and which can now be eliminated or minimised. In the control phase, the researchers will thus be able to understand what aspects of marketing spending yield the highest returns.
The following tools were utilised with a DOE application as outlined in Tools utilised:
Tools utilised
Problem statement/project charter.
Data collection, DOE and correlation analysis.
DOE analysis – factorial regression, normal plot, Pareto charts and brainstorming.
DOE response optimisation/brainstorming.
Response plot.
Source: Authors’ own work
4. Results
4.1 Planning of the experiment
As stated previously, the problem statement is that a lack of utilisation, nonvalue-add steps and poor process set-up of safelist mailers and traffic exchanges websites is noticed within the Julsine store. This meant fewer hits and interactions with products from a marketing point of view, thus decreasing revenue opportunities. Therefore, this research aimed to improve the digital marketing process using LSS methods to resolve the issues.
Firstly, a DOE was carried out to aid control, predict and optimise a complex process for any desired behaviours (Montgomery, 2014). Then, three safelist mailers, four traffic exchanges and time were chosen as factors that impact the hits and visits, with two levels determined. A level is a value that a process variable holds in an experiment, e.g. a car’s gas mileage is influenced by such levels as tyre pressure, speed, etc. (Antony, 2014). These factors and levels were set up as follows in Table 1.
Factors for DOE
| A. | Factors |
| (1) | TE |
| – | Easyhits4U.com (EHU) |
| – | Memberrules.com (MR.) |
| – | Free Traffic Lotto.com (FTL) |
| – | MoneyMakerXchange.com (MMX) |
| (2) | SM |
| – | Europeansafelist.com (ES.) |
| – | Adchiever.com (AC.) |
| – | MailOurList.com (MOL) |
| (3) | TIME |
| B. | Levels |
| (1) | For websites |
| – | 200 credits (−) to allocate for each experiment |
| – | 800 credits (+) to allocate for each experiment |
| (2) | For Time |
| – | 24 hours (−) |
| – | 48 hours (+) |
4.2 Design of the experiment
A full factorial design for studying eight factors at two levels would have required 256 experimental trials (2n=8). The twice replication would have taken the total number of observations to 512. To reduce the number of experimental trials while maintaining statistical integrity and providing statistically valid results, the fractional factorial design 28–4 IV was selected from the array of designs popularised by the classical DOE (Antony, 2006; Fisher, 1960). Thus, the total number of experimental trials was reduced from 512 to 32, from which a fractional factorial was designed. Figure 1 below represents the fractional design 28–4 IV planned and created using Minitab before the experiment.
4.3 Conducting the experiment
One operator carried out the entire experiment to minimise operator-to-operator variability. Moreover, the researcher used the randomisation strategy to minimise noise factors’ effect, which causes variability in performance. The blocking strategy would be used for the same reason, but it was not. Digital marketing is one platform that operates in different areas simultaneously and worldwide. Thus, the hits and visits come from all over the world.
Two replicates of a 28–4 fractional factorial design were performed on a Digital Marketing process of the Julsine Online Store. Two responses, hits and visits were measured using the internet and site store at each treatment combination. The results are shown in Table 2 below.
Results of measurement
| # | EHU | MR | FTL | MMX | ES | AC | MOL | TIME | HITS | VISITS | ||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| I | II | I | II | |||||||||
| 1 | −1 | −1 | −1 | −1 | −1 | −1 | −1 | −1 | 49 | 75 | 46 | 72 |
| 2 | 1 | −1 | −1 | −1 | −1 | 1 | 1 | 1 | 29 | 100 | 26 | 74 |
| 3 | −1 | 1 | −1 | −1 | 1 | −1 | 1 | 1 | 287 | 310 | 285 | 300 |
| 4 | 1 | 1 | −1 | −1 | 1 | 1 | −1 | −1 | 339 | 304 | 330 | 297 |
| 5 | −1 | −1 | 1 | −1 | 1 | 1 | 1 | −1 | 116 | 230 | 115 | 192 |
| 6 | 1 | −1 | 1 | −1 | 1 | −1 | −1 | 1 | 20 | 90 | 20 | 74 |
| 7 | −1 | 1 | 1 | −1 | −1 | 1 | −1 | 1 | 66 | 70 | 53 | 65 |
| 8 | 1 | 1 | 1 | −1 | −1 | 1 | 1 | −1 | 130 | 252 | 110 | 215 |
| 9 | −1 | −1 | −1 | 1 | 1 | 1 | −1 | 1 | 87 | 147 | 80 | 130 |
| 10 | 1 | −1 | −1 | 1 | 1 | −1 | 1 | −1 | 160 | 270 | 148 | 200 |
| 11 | −1 | 1 | −1 | 1 | −1 | 1 | 1 | −1 | 78 | 99 | 71 | 92 |
| 12 | 1 | −1 | 1 | −1 | 1 | −1 | −1 | 1 | 121 | 172 | 115 | 120 |
| 13 | −1 | −1 | 1 | 1 | −1 | −1 | 1 | 1 | 73 | 70 | 63 | 70 |
| 14 | 1 | −1 | 1 | 1 | −1 | 1 | 1 | −1 | 31 | 81 | 21 | 64 |
| 15 | −1 | 1 | 1 | 1 | 1 | 1 | −1 | −1 | 60 | 120 | 59 | 103 |
| 16 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 225 | 350 | 220 | 217 |
4.4 Analysis of results of experiments
This research aims to analyse the impact of factors and interactions on the hits and visits to the store site. A model including factorial regression, normal plot and Pareto charts of the standardised effects has been constructed using Minitab to achieve this objective.
A model was constructed for each response and refined to include only the statistically significant terms and the terms needed to maintain the hierarchy. Therefore, the model for hits was reduced with the terms E, B, G, AB, AH, A, C, AD, D, F and H, and the model for visits was refined with the terms E, B, AH, G, AB, C, D, A, AF, AD, AG, F and H (Figure 2).
The analysis of variance (ANOVA) table, the normal plot and the Pareto chart of the standardised effects indicate that E, B, G, AB, AH, A, C and AD are statistically significant at a 5% level.
The model summary shows that the adjusted R2 is 75.81%, and R2 predict 60.04%. Therefore, the adjusted R2 indicates that the model has accounted for 75.81% of the variation that is observed in the data:
The graphical tools, namely, the normal plot and the Pareto chart of the standardised effects, were used to support and justify this claim.
The normal plot (Figure 3) shows that the terms E (ES), B (MR), G (MOL), AB (EHU*MR), AH (EHU*TIME), A (EHU), C (FTL) and AD (EHU*MMX) are active and statistically significant and fall off the straight line whereas the inactive and non-significant fall along the straight line (Conover, 1980).
The Pareto chart of the standardised effects below (Figure 4) shows that the terms E (ES), B (MR), G (MOL), AB (EHU*MR), AH (EHU*TIME), A (EHU), C (FTL) and AD (EHU*MMX) are statistically significant because they exceed the threshold line 2.086 (Fisher, 1960).
This HITS model’s residual plots indicate that ANOVA assumptions of normality, equal variance and independence have not been violated. Thus, the transformation of the model is not needed. The results of the factorial regression as a model for visits are shown in Figure 5.
The ANOVA table, the normal plot and the Pareto chart of the standardised effects indicate that E, B, AH., AB, C, G, D AF, A and AD are statistically significant 5% level.
The model summary shows that the Adjusted R2 is 87.81%, and R2 predicts 77.64%. Therefore, the adjusted R2 indicates that the model has accounted for 84.64% of the observed variation in the data:
To determine the non-violation of the ANOVA assumptions of normality, equal variance and independence, the residual plots graphs for visits, including the non-significant terms, have been generated (Figure 6).
The selection of optimal settings depends on the experiment’s objective or the nature of the problem under study. This study aims to analyse the data from hits and visits responses to achieve a process setting that optimises these two responses. As this study deals with two responses and more than three main factors, the suitable techniques are the desirability function (Vera Candioti et al., 2014). Using the Minitab inbuilt response function, the response optimisation table and the optimisation plot below have been generated (Figures 7 and 8).
The individual desirability for HITS is 1.0, meaning that the response HITS has been minimised to 300. The individual desirability for VISITS is 0.97416, and the overall desirability or composite desirability for both responses is 0.986995 (0.98700). The objective is to make these two responses closer.
This desirability is achieved when the optimal factors setting is as follows: A: −1, B: 1, C: −1, D: −1, E: 0.996294, F: −1, G: 1, and H: 1.
This setting has been translated from coded to uncoded levels to make the correct adjustment to the process.
+1: 800
–1: 200
0.996294: 797
The DOE also helped implementation of a 5S program as far as the results of the analysis informed the website Marketing manager as to what they could “tidy up” in terms of the online site configurations and “declutter” the site. The research also shows that 5S can be utilised in online site housekeeping and organisation as far as it makes it possible to rid the Marketing process of unnecessary “inventory” waste via excess website material and layouts and thus avoids “overprocessing” wastes. In tangent with the results from the DOE analysis, a 5S process was implemented to ensure the site was more visual and thus ease organisation of configurations.
The first step in the implementation is predominantly based on the control approaches to confirm the optimum setting’s effectiveness (Shruti, 2018). Before onward actions regarding implementation, a three-time confirmatory experiment was undertaken using the optimal factors setting (Table 3). As a result, the store site recorded an average of 299 hits and 293 visits.
Results of confirmatory run
| # | EHU | MR | FTL | MMX | ES | AC | MOL | TIME | HITS | VISITS |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 200 | 800 | 200 | 200 | 797 | 200 | 800 | 48 | 298 | 289 |
| 2 | 200 | 800 | 200 | 200 | 797 | 200 | 800 | 48 | 297 | 294 |
| 3 | 200 | 800 | 200 | 200 | 797 | 200 | 800 | 48 | 301 | 295 |
| Average | 299 | 293 | ||||||||
This confirmatory experiment shows an improvement of 99% in the average hits and 144% in the average visits from before the experiment. In addition, the time spent on marketing has been reduced from over 60–90 min a day down to 30 min daily. Following Bala and Verma (2018), every digital marketing campaign aims for a positive return on investment but not every campaign leads directly to a profit increase. This study increased the likelihood of profit by removing non-value add configurations.
Two categories of control methods are used to know if the digital marketing efforts succeeded based on the outcome of the DOE. These are the conversions and the website’s behaviour. Conversions occur when visitors convert their visits into purchases. The website behaviour aims to determine the digital campaign success based on the customers’ visits to the store site (Schriffman and Kanuk, 1994).
The first method of control was analysing the website visits from the website behaviour category used to measure the visits regarding each digital campaign. The second is the online sales from the conversions category, which will be used to track online sales. Unfortunately, these control methods are not preventive as they only allow spotting potential problems early on if a sudden drop in visits and/or sales occurs, but they do allow an early indicator of these problems.
5. Discussion
Marketers are facing a marketplace increasingly more specialised, more globalised and more technologically driven (Jackson, 2022). Products are increasingly desired and wanted by customers. Marketing continually strives to gain a niche and increase gain market share. Design of experiments is well-established a protocol for designing, developing and improving products and processes, and thus can be used to help development and production to achieve marketing’s goals (Frey and Haller, 2021). This study demonstrated how the online store could operate more efficiently from a marketing viewpoint and improve process efficiency in a very resource constrained online website. The level of information gained from following the DOE protocol and the continual improvement steps provided what is important to the customer and aided in removing inefficiencies (Boyles, 2010).
Specifically, in this study, two types of digital marketing, namely, safelist mailer and traffic exchange websites, were discussed. The proven application of design of experiments (DoE) has demonstrated the tools evolution as a powerful industrial statistics tool for managing and improving processes in diverse sectors and processes (Prashar, 2016). Thus this study aided marketing in adopting a new paradigm for them to improve their customer experience (Frey and Haller, 2021).
The DOE can be applied to other auto-surf, auto view or auto-clicks websites. Many websites are manual and are not performing in terms of generating hits and visits. The processes that use them waste valuable time and the cost of the credits needed for the promotion unnecessarily increase the price of the products and/or services. The consequence is that statistics only show the hits or clicks without effective visits. DOE application can highlight this non-performance.
Thus, this study has met the RQ to ascertain if DOE methodology can help improve the quality of the Digital Marketing process. As a Lean Six Sigma and process optimisation tool, DOE is an appropriate tool and strategy for improvement in this case as demonstrated in many studies (Antony, 2014; Starkey et al., 1997).
Specifically, the following was identified:
DOE application identified the safelist mailer and traffic exchange websites with a higher number of hits and visits to an online store and minimise the time and the cost of products and services. It is common knowledge that the higher the number of visitors to an online store, the greater the likelihood of increasing sales volumes (Adam et al., 2020). The strategy here consists of identifying performing websites and eliminating or reducing work and focus on nonperforming ones. The DOE usage and application allowed the marketing manager to identify the performing websites and nonperforming ones thus eliminating waste and optimising resources (Antony, 2006).
Minimized the cost of credits and maximised the use of credits needed for the promotion and avoid the presence of unused credits that can unnecessarily increase the price of products and services. Any amount spent to purchase credits constitutes a charge that is incorporated into the cost of the promoted products and services. In other words, DOE allowed the case study organisation to determine the number of credits to allocate to each website during a specific period.
Following, every digital marketing campaign aims for a positive return on investment. Not every campaign leads directly to the profits increase. Some raise business awareness and others generate visits to the store. Both aim that profits increase in a long run. Short-term results cannot be measured with monetary values. However, the website was able to show a time and labour cost reduction (Before versus after DOE was applied).
To control the improvement, two methods were used, namely, the conversion and the website behaviour. The conversion occurs when visitors convert their visits into purchases and the website’s behaviour aims to determine the campaign’s success based on the customers’ visits to the store. However, these controls are not preventive. They only allow the spotting of potential problems early on if a sudden drop in visits and or sales occurs. Thus DOE in this case aided in visual management and acted as a type of error proofing or poke-yoke (Steenkamp et al., 2017).
The research shows that DOE methods can help improve digital marketing quality by generating more interactions with safelist mailers and traffic exchange websites, reducing time and money spent on unnecessary marketing efforts, and maximising the use of credits needed for promotion. The use of DOE has been demonstrated for process improvement in manufacturing environments where there is a need to evaluate the factors that control the value of a parameter or group of parameters (Bower, 2023). This study has expanded the use of traditional DOE applications from manufacturing into marketing and is unique in this respect.
The project was deemed successful by the store with careful management of efficient websites’ distinction from those that do not generate enough traffic. This overcomes a problem highlighted by many digital marketing studies (American Marketing Association, 2022; Bala and Verma, 2018; Idrysheva et al., 2019).
From an operational results viewpoint, the research highlighted the terms E, B, G, AB, AH, A, C, AD, D, F and H as being positively significant for the response Hits and the terms E, B, AH, G, AB, C, D, A, AF, AD, AG, F and H as being positively significant for the response Visits. This means that they generated enough traffic needed by the store. On the other hand, the terms D, F and H are deemed not significant or negatively significant for Hits and F and H for Visits. This means that the membership subscription fees and the credits regarding these terms are not used or partially used and unnecessarily increase the products’ price. Therefore, they can be considered as waste to eliminate from the process (Antony, 2014; Dixon et al., 2006).
Any expense incurred in producing, procuring and promoting products for sale is reflected in those products’ market prices. Therefore, implementing DOE within a Six Sigma initiative has to start by considering what the customers require and the voice of the customer (George, 2002; McDermott and Antony, 2021). One of the basic tenets of customer satisfaction and quality management is to offer low product prices to customers (Snee and Hoerl, 2002). This research delivered via DOE application has meant that product price increases can be avoided by eliminating unused credits and membership subscription fees related to inefficient Digital channels, which are non-value added.
Thus, DOE application allowed the store to allocate appropriate credits to Digital channels deemed efficient and avoid paying membership subscription fees to those that do not work correctly. Consequently, the time spent on the marketing process was reduced, giving a productivity saving of approx. 1 h daily. This application thus delivered process optimisation and enhanced resource utilisation (Antony et al., 2024; Besseris, 2011).
The Lean tool 5S was utilised to improve flow and layout and enhanced visibility on the site, as it has been demonstrated to improve physical layout in manufacturing (Nelson et al., 2022; Trubetskaya et al., 2022). Few documented studies of 5S in an online environment exist, but Nelson et al. (2022) have highlighted its use in the online administration and invoicing processes of micro enterprises.
Within case analysis demonstrates that a reduction in unnecessary work elements and non-value add waste improved morale within the marketing store as time was optimised. Also within case demonstrated the differences between how different websites worked and thus how credits were allocated and an awareness of how things could be completed differently and allows us develop the research propositions that follow (Ayres et al., 2003).
The additional advantage of this project can be the increased awareness of the impact of digital marketing and the potential use of such a strategy in any digital marketing using safelist mailers and traffic exchange websites. The Julsine store will expand this type of experiment to detect the most efficient safelist mailers and traffic exchange websites on the internet and generate a series of optimum settings for improving its marketing process.
Based on our findings we believe the benefits of applying DOE within a DMAIC process framework in a marketing online store was effective. A high-level quantitative result comparison of before versus after the DOE application is highlighted in Before versus after results. The case study approach aided in the communication of a real cases and application of DOE in a Marketing environment providing a rich source of information from which academics and practitioners “can study the challenges of applying and deploying theory in a complex and messy environment of a real organization” (Lameijer et al., 2024):
Before versus after DOE was applied
99% increase in average hits to the website.
144% increase in average visits to the website.
Reduction of 60–90 min a day from time spent on marketing activity.
Elimination of unnecessary or poor-performing configurations/decluttering of the website.
Elimination of “inventory” related to website excess material.
More visual, easier-to-manage website.
Source: Authors’ own work
6. Conclusion, implications, limitations and directions for further research
The role of marketing is to promote products that customers covet and gain market share. This paper has met the research objectives to ascertain if DOE methodology can help improve the quality of digital marketing (RQ). In addition, the project generated more interactions with safelist mailers and traffic exchanges websites, thus minimising time spent in the marketing process and maximising the use of the credits needed for promotion. The efficiency of DOE means that marketeers can leverage this in designing processes in resource-constrained marketing functions. The depth of information gained from following the DOE steps and the continuous improvement cycle provides the marketeer with what is important to the customer. The study has both theoretical and practical implications. From a practical viewpoint, marketing has been slow to embrace tools such as DOE, and this paper proves that DOE can be utilised as a valuable continuous improvement tool, by enhancing marketing processes and making them more responsive. The study demonstrated that the embracing of tools like DOE allows the marketing function to optimise productivity and time and utilise tools that are consistent with the voice of the customer. Thus, DOE offers a new departure or paradigm for marketeers. Also, this study enhances the state-of-the-art literature and fills a research gap by deploying DOE to online marketing stores and proves that DOE tools can be deployed further in other marketing functions thus aiding the pedagogical value to the field of marketing. Further the study has implications to increase the knowledge and application of the impact of DOE on digital marketing as a strategic tool in using safelist mailers and traffic exchange websites to increase online marketing store interactions. DOE application can aid the identification of online site wastage and optimise the site’s efficiency, generating hits and visits, thus increasing sales and revenue opportunities.
While a limitation of the study is that it is a single case study, applying DOE to an online marketing store is generalisable for other marketing organisations and offers a new approach to delight the customer. Leveraging DOE allows marketing to enhance and shape management practice, but also to embrace an opportunity to create and enhance a toolkit that is consonant with marketing’s underlying philosophy and orientation.
Future research opportunities will be to deploy DOE in other areas of marketing processes to demonstrate the applicability of statistical analysis and DOE to those processes. Also, this case study analysis should be carried out in other online stores to compare learnings and opportunities.
