Shopping marketplace analysis based on customer insights using social media analytics

Abstract

Using the recent advances in data analytics, Companies can leverage sentiment analysis to identify trends and areas for improvement of their market strategy and operation planning. This analysis allows them to understand the sentiments expressed by customers and accurately predict customer behavior. Social media platforms have fundamentally altered the field of digital marketing. The findings of this research try to provide:

• •Potential implications for businesses aiming to optimize their product offerings and enhance customer satisfaction within specific cultural contexts.
• •A case study has been designed for understanding customers' perceptions and satisfaction levels toward various shops, including local ethnic stores and big chain stores.

The paper has conducted a literature review around main pillars like application of data analytics on social media, and sales strategies for establishing a marketplace using data analytics. Exploring the utilization of social media and customer feedback, the paper has proposed a conceptual model for creating insights for extraction of shopping experiences and factors used for customer purchasing decision making.

Graphical abstract

Specifications table

Subject area:	Engineering
More specific subject area:	Social Media Data Analytics
Name of your method:	Customer perception for products and services using social media analytics
Name and reference of original method:	Maurya, J., Goyal, K., Misra, A.K., 2022. Modeling the impact of social media on the adoption of a new product by customers. MethodsX 9, 101932. https://doi.org/10.1016/j.mex.2022.101932; Mohammadian N, Salman S, Uygun Y, Fatahi Valilai O. Embedding Perceptual Quality in Omnichannel's Touchpoints in Product Development Lifecycle Management Using Data Analytics. In: Silva FJG, Ferreira LP, Sá JC, Pereira MT, Pinto CMA, editors. Flexible Automation and Intelligent Manufacturing: Establishing Bridges for More Sustainable Manufacturing Systems, Cham: Springer Nature Switzerland; 2024, p. 999–1010. https://doi.org/10.1007/978-3-031-38165-2_115.
Resource availability:	The data that support the findings of this study are available from the corresponding author upon reasonable request

Background

In today's highly competitive business landscape, companies that seek to thrive must understand and influence customer perspectives [1]. The advent of advanced technologies has given companies powerful tools to accomplish this. Sentiment analysis, social media, Industry 4.0 and simulation techniques such as digital twins have emerged as effective means for companies to shape and impact customer perspectives and their potential marketplaces [2]. By utilizing these powerful tools, businesses can gain valuable insights into customer sentiment and factors that affect establishing a marketplace for products, actively engage with them on social platforms, and simulate customer behaviors in virtual environments [3].

Social media platforms have fundamentally altered the field of digital marketing. These online venues facilitate open dialogue among participants, allowing them to exchange viewpoints, recount personal experiences, and engage in discourse on assorted matters. Marketing professionals capitalize on these opportunities to expand their customer base, obtain feedback, and form durable ties with patrons [4,5].

In today's globalized world, with a significant increase in immigration, cities across the globe host diverse populations representing various nationalities. This phenomenon has given rise to various stores that specialize in selling local products from different countries. By leveraging social media data, businesses can gain valuable insights into where their target customers prefer to shop and subsequently make informed decisions regarding product placement and marketing strategies. The findings of this research try to provide potential implications for businesses aiming to optimize their product offerings and enhance customer satisfaction within specific cultural contexts.

The main motivations for using data analytics on social media can be discussed as:

• •Analyzing user-generated content and opinions on social media platforms like Google.
• •Assisting businesses in understanding the underlying dynamics of marketplace choices.
• •Findings can inform marketing strategies and enhance customer targeting.

formats, is rapidly generated on different platforms based on social media [6]. A report from the exploding topics website suggests an astounding daily creation of approximately 328.77 million terabytes of data [7]. This remarkable data surge can be attributed to the platform's ability to facilitate swift information sharing, support various media types like text, images, audio, and videos, and enable users to interact with many peers simultaneously. Given these characteristics, social media emerges as a promising data source that directly reflects the opinions and perspectives of consumers on numerous subjects [8]. Data analytics serves as a valuable tool to reveal compelling narratives from this vast sea of data, making information more lucid and accessible. These advancements enable organizations to extract valuable insights and make data-driven decisions, enhancing efficiency and effectiveness across industries. With the power of Big Data analytics, businesses, healthcare, finance, and other sectors have seen significant improvements in performance and innovation.

Investigation of the influence of distribution channel strategies on consumer behavior using survey methods and quantitative analysis has been considered by recent studies. Wei and Dong [9] Focused on online retail platforms, this research examines the interaction between suppliers and different distribution strategies. In comparison, Rangan et al. [10] presented a method for selecting distribution channels for new products through expert evaluations and analytical synthesis. Shen et al. [11] in comparison, investigates the interaction between a manufacturer, platform retailer, and traditional reseller in China's electronics market context.

Sentiment analysis, also known as opinion mining, is a computational technique that involves analyzing textual data to determine the sentiment expressed, classifying it as positive, negative, or neutral [12]. It leverages natural language processing and machine learning methods to extract subjective information from text, providing valuable insights into customer sentiments, brand reputation, and consumer opinions. In the context of product development, sentiment analysis plays a crucial role in understanding customer feedback and opinions about products or services [13]. Sentiment analysis involves various important stages, including data preprocessing, feature extraction, model training, and sentiment labeling [14,15]. In conclusion, sentiment analysis plays a vital role in the product development process by enabling organizations to gain a deeper understanding of customer sentiments and opinions. By analyzing textual data, sentiment analysis provides valuable insights into customer satisfaction, preferences, and areas of improvement [16,17].

This paper aims to explore the utilization of social media and customer feedback as valuable resources to gain insights into their shopping experiences and the factors they prioritize while making purchasing decisions. Employing sentiment analysis on the extensive data available on social media platforms, companies can create a digital twin representation of real markets. This twin offers a unique opportunity to comprehend customer sentiments, preferences, and needs in real-time, facilitating a comprehensive understanding of consumer behavior. The study delves into the sale strategy aspect for companies, examining the potential of using social media and sentiment analysis to effectively select suitable stores for their products. By evaluating customers' comments, reviews, and sentiments towards various retail outlets, companies can pinpoint the stores that align with their target market, brand values, and product offerings. By employing data analysis, businesses can confidently determine ideal shop locations and fine-tune their marketing tactics according to shopper opinions and inclinations.

Method details

This paper aims to investigate the sentiments expressed in Google reviews of selected supermarkets in two types of Ethnic Stores and Chain Stores in Germany through the application of sentiment analysis. Ethnic stores focus on providing products authentic to specific cultural communities, whereas chain stores have a broader array of standardized goods with an emphasis on convenience and cost-effectiveness. While some chain stores may have sections dedicated to ethnic products, their primary focus is on mass-market goods. Ethnic stores, on the other hand, have their main emphasis on providing culturally authentic products for specific communities, although some may offer sections for mainstream items.

As shown in Fig. 1 The relationship between users in social media, stores (such as local ethnic shops or chain stores), and marketing departments is a complex and interconnected ecosystem that revolves around customer experiences and market insights. Customers who visit local shops or chain stores to purchase groceries or other products eventually become social media users. These users often share their experiences, opinions, and feedback about their shopping encounters on social media apps, creating valuable user-generated content. Marketing departments can harness the data generated through these user interactions on social media to gain deeper insights into customer sentiments and preferences. With the help of AI and sentiment analysis, marketing departments can analyze users' feedback and sentiments shared on social media platforms. This analysis enables companies to understand customers' perceptions and satisfaction levels toward various shops, including local ethnic stores and big chain stores.

Fig. 1.

Proposed conceptual model framework.

By implementing digital twin technologies, marketing departments can create virtual representations of actual markets, where the behavior and preferences of users in social media are simulated in real-time. This digital twin approach provides a comprehensive understanding of customer behavior and trends, allowing companies to identify potential target markets that align with their product offerings. With these insights, marketing departments can make data-driven decisions regarding their marketing or sales strategies. They can choose the best markets or store locations to sell their products, ensuring that they reach their target market effectively. Furthermore, this approach helps companies tailor their marketing strategies to cater to specific customer preferences and demands, ultimately enhancing customer satisfaction and brand loyalty. In summary, the relationship between users in social media, stores (local and chain), and marketing departments revolves around data analysis, user feedback, and market insights. Through the use of AI, sentiment analysis, and digital twin technologies, marketing departments can optimize their strategies and target the right markets to position their products effectively. This interconnected ecosystem ensures that customer experiences, user feedback, and marketing efforts are aligned, driving business success in a competitive marketplace.

This study encompasses an analysis of 31 supermarkets, comprising 18 Persian Ethnic food stores, 5 Arabic Ethnic food stores, 5 Turkish Ethnic food stores, and 8 German Chain stores. In total 3777 comments have been made over a period of five years for this study. In the initial phase of this research, the data collection process involved extracting Google reviews utilizing web scraping techniques. Web scraping is a methodology that entails automatically retrieving information from websites by parsing their HTML structure [18]. In this case, the focus was on gathering Google reviews, which are user-generated feedback and opinions about various supermarkets across Germany. Moving forward, the process includes recognizing the sentiment classification and emotion models utilized. The initial one is Vader (Valence Aware Dictionary for Sentiment Reasoning), adept at gauging polarity and emotional intensity. Vader operates as a rule-based sentiment analysis tool tailored for social media expressions. Notably, it's both free and open source, delivering scores for positivity, negativity, and neutral along with comprehensive sentiment insights. To derive a text's sentiment score, one can compute the product of individual word intensities within the text [19,20]. Secondly, Flair, a natural language processing library (NLP) created and open-sourced by Zalando Research, employs binary sentiment classification to determine whether sentiment is negative or positive, alongside a confidence coefficient on a scale from 0 to 1. In contrast to other NLP toolkits that rely on lexicons, Flair's sentiment classifier utilizes a character-level LSTM neural network. This network considers both sequences of letters and words to classify text based on the corpus. The pre-trained sentiment analysis model within Flair has been trained on the IMDB dataset and exclusively assigns either a positive or negative polarity to texts [21]. The sentiment and emotional facets of tweets are analyzed through the utilization of the Robustly Optimized BERT Pretraining Approach (RoBERTa). RoBERTa [22] stands out as an enhanced iteration of the Bidirectional Encoder Representations from the Transformers (BERT) model, marked by extended training epochs, larger training batches, and an augmented training dataset. Serving as the third model within this paper, Twitter-RoBERTa-base is employed to classify the reviews' emotions and polarity. This model is pre-trained on a Twitter dataset encompassing over 58 million tweets [23].

Method validation

To begin the analysis, the collected Google reviews were processed and the most frequently occurring words were identified using the Word Cloud technique. Fig. 2(a) displays the word cloud representing ethnic stores, while Fig. 2(b) showcases the word cloud for chain stores, providing a visual representation of the prevalent terms in the reviews. Word size is contingent upon frequency or significance in the text, accentuating salient terms. This approach offers a hierarchical view of textual prominence [24,25].

Fig. 2.

Word Cloud visualization of data for (a) Ethnic stores, (b) Chain stores.

Notably, both visualizations reveal a consistent prevalence of the term "good," which is prominently repeated across reviews. In the word cloud associated with ethnic stores, frequently used words include "product," "price," "good," "nice," "shop," "fresh," and "friendly." Conversely, in the word cloud of chain stores, frequently recurring words comprise "always", "staff", "good", "store", "great", "small", "friendly", and "REWE." This word cloud analysis effectively highlights the most commonly employed positive and negative terms. Ethnic stores are applauded for their friendly staff and welcoming ambiance, creating a personalized experience. Customers appreciate products' excellent quality and affordability, mainly fresh fruits, vegetables, and meat. Diverse and well-curated selections cater to different preferences, while the availability of halal items addresses specific dietary needs. However, concerns arise about occasional high prices, expired products, and rare instances of rude staff behavior. Chain stores stand out for their extensive product offerings, helpful staff, and discounts.

These stores provide convenience with ample parking and organized layouts, making shopping efficient. Despite positive attributes, some customers note overcrowding, cleanliness issues, and long checkout queues, impacting the overall experience. Additionally, mixed feedback exists about complex store layouts, overpriced items, and occasional inefficiencies at checkout. This analysis provides valuable insights for suppliers seeking to optimize product placement based on customer feedback. By understanding customer sentiments and preferences, suppliers can strategically position their products in ethnic stores and chain stores to maximize sales and customer satisfaction.

Furthermore, Fig. 3 showcases the count number of emotions identified for each group of markets, providing a comprehensive overview of the emotional responses expressed by customers towards the supermarkets, enabling a comparative analysis of emotional responses across the different markets. Figures 5 and 6 are created by PowerBI©. The emotion analysis of Google reviews was performed on both ethnic stores and chain stores in Germany, utilizing the Twitter-roBERTa-base model. The objective was to extract the underlying emotional tones prevalent in the reviews and comments for these two distinct categories of stores. The analysis encompassed a total of 1931 comments for ethnic stores and 1846 comments for chain stores.

Fig. 3.

Emotion analysis for different ethnic and chain stores.

Across both ethnic stores and chain stores, a common thread of positivity was evident. Joy, optimism, love, and trust were consistently expressed in the reviews for both categories. Customers of ethnic stores appeared to be especially engaged, with a higher frequency of comments conveying joy, optimism, love, and trust. This suggests that ethnic stores managed to create an environment that fostered positive emotional connections with their customers. While positive emotions were prevalent, negative emotions were also present in the reviews, albeit to different extents. Ethnic stores exhibited a lower number of comments expressing anger and disgust compared to chain stores. This implies that customers of ethnic stores were less likely to voice strong negative sentiments. On the other hand, chain stores received a relatively higher number of comments indicating anger and disgust, indicating potential areas of dissatisfaction among their customers. The analysis further revealed nuanced emotional responses in the reviews. Customers of ethnic stores frequently mentioned emotions like anticipation and excitement, which could be attributed to the uniqueness of the products or experiences offered. Similarly, chain stores evoked emotions of anticipation, potentially linked to the expectations associated with well-known brands.

Moreover, the sentiment analysis process involved the application of three pre-trained NLP model, "Twitter-roBERTa-base". This model was used to classify the reviews into positive, negative, or neutral sentiment categories. Fig. 4 displays the sentiment classification of Ethnic supermarkets based on different models and the classification distribution for Chain stores. In the dynamic landscape of a marketplace, where suppliers strive to optimize their product placements, the ability to discern the sentiment of comments becomes a powerful tool. This sentiment analysis of user feedback paints a compelling picture: ethnic stores emerge as favorable platforms for product distribution. The data underscores a higher percentage of positive comments in ethnic stores compared to their chain store counterparts. This positive sentiment not only indicates customer satisfaction but also reflects the potential for greater engagement and loyalty. Equally significant is the revelation that ethnic stores experience a notably lower percentage of negative comments in contrast to chain stores. This disparity suggests a more harmonious shopping experience and better customer interactions within the ethnic store environment. Suppliers, armed with this knowledge, gain a strategic advantage by opting to channel their products through ethnic stores. The positive sentiment prevailing in such establishments can serve as a persuasive marketing tool, while the reduced negativity assures a more conducive atmosphere for brand promotion and customer engagement.

Fig. 4.

Sentiment classification of (a) Ethnic stores (b) Chain stores based on the Twitter-roBERTa-base model.

In conclusion, this paper has tried to investigate the customer experiences and preferences when selecting shopping markets by leveraging the power of social media data analysis. Through the meticulous examination of social media content, this study has shed light on the intricate dynamics that guide customers' decision-making processes.

The utilization of social media data analysis presents a valuable opportunity for companies and suppliers to gain profound insights into customer experiences and preferences. By tapping into the wealth of information shared by consumers on various social platforms, businesses can uncover invaluable feedback that might have otherwise gone unnoticed. This feedback can then be harnessed to refine and optimize the shopping experience, tailoring offerings to align more closely with customer expectations. Emotion analysis, a significant facet of this study, has underscored the significance of positive emotions in shaping customer perceptions. Notably, the sentiment analysis has revealed that chain stores could benefit from implementing strategic marketing initiatives aimed at enhancing positive emotional responses. This sentiment analysis has the potential to serve as the foundation for constructing digital twins of real-world stores. To extend the groundwork, a longitudinal analysis could be undertaken to track the evolution of customer experiences and preferences over an extended timeframe. This would provide dynamic insights into how these factors shift with changing societal trends and technological advancements.

Credit author statement

Negar Nasrabadi: Conceptualization, Methodology, Software, Writing - Original draft preparation. Hendro Wicaksono: Validity check, Data curation, Reviewing and Editing. Omid Fatahi Valilai: Visualization, Methodology, Investigation, Supervision, Validation, Writing- Reviewing and Editing.

Limitations

None.

Ethics statements

This study involves data collected from social media platforms, the authors confirm that a) informed consent was obtained from participants or that participant data has been fully anonymized, and b) the platform(s)’ data redistribution policies were complied with.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

• Data will be made available on request.