Black Friday represents a peak in the retail calendar. Millions of consumers wait for this specific window to make significant purchases, particularly in the technology sector. While sales figures often dominate the headlines, a parallel narrative unfolds on social media where customers express their unfiltered reactions to these promotions. This digital chatter offers a window into consumer psychology that goes beyond simple transaction data.
Researchers from Rey Juan Carlos University and the University of Seville in Spain viewed this online conversation as a source of business intelligence. They sought to understand how consumers truly perceive the marketing strategies employed by major technology companies during these high-stakes sales events.
Their investigation, published in the Journal of Open Innovation: Technology, Market, and Complexity, applied data mining techniques to analyze user sentiment. The team aimed to determine which promotional tactics foster engagement and which ones provoke public backlash.
The Question of Consumer Trust
The research team, led by Jose Ramon Saura, Ana Reyes-Menendez, and Pedro Palos-Sanchez, identified a growing tension in the retail landscape. While Black Friday participation remains high, skepticism regarding the veracity of offers has increased. A common consumer complaint involves companies raising prices weeks before the event, only to lower them back to the original price while labeling it a discount.
The researchers wanted to know how these perceptions manifested in public discourse. They focused on the concept of User-Generated Content (UGC). This term refers to any form of content, such as tweets or reviews, created by unpaid contributors. The team operated under the premise that this content summarizes key information that influences the purchasing decisions of future customers.
To explore this, the researchers designed a study to analyze interactions between users and the twenty-three largest technology companies in Spain. They specifically looked for patterns in how users categorized their experiences. The goal was to move beyond general sentiment and identify the specific topics that drove positive or negative reactions.
Mining the Data
The investigation focused on the Black Friday period of 2018. The team collected data over a seven-day window that included the three days leading up to the event, the day itself, and the three days following. They utilized the Twitter API to download 5,064 tweets that contained the hashtag #BlackFriday and interactions with the selected companies.
The researchers then refined this dataset. They removed repeated tweets to ensure unique data points. They also excluded tweets that contained only images or videos, as the study focused strictly on Natural Language Processing. Retweets were treated as independent data points because they represented an endorsement by a unique user.
A Three-Step Analytical Process
The team employed a three-stage methodology to process the textual data. The first step involved a technique called Latent Dirichlet Allocation (LDA). This is a mathematical model used to discover abstract topics within a collection of documents. The algorithm scanned the database of tweets to measure word frequency and recurrence. It then grouped these words into distinct clusters, which the researchers labeled as topics.
The second step applied a sentiment analysis algorithm to these identified topics. This machine learning tool classified the content associated with each topic into three categories: positive, negative, or neutral. To ensure the computer explicitly understood the human emotion in the text, the researchers calculated a reliability score known as Krippendorff’s alpha value.
The third step involved textual analysis. This process allowed the researchers to examine the specific context within the data nodes. By manually reviewing the information classifiers, they identified the underlying reasons for the sentiment scores. This revealed the specific business practices that triggered user responses.
Categorizing the Conversation
The LDA model divided the user comments into seven distinct topics. These included “Offers and Discounts,” “Exclusive Promotions,” “Fraud,” “Insults and Noise,” “Smartphones,” “Computers Accessories,” and “Customer Support.” The sentiment analysis revealed a stark contrast in how users perceived these different categories.
Two topics consistently generated positive feelings: “Exclusive Promotions” and “Smartphones.” The textual analysis showed that users reacted favorably to offers they perceived as genuine and time-limited. Promotions that lasted only for hours or minutes created a sense of exclusivity that users appreciated.
In contrast, three topics drew significantly negative reactions: “Fraud,” “Insults and Noise,” and “Customer Support.” The data showed that users utilized Twitter to call out perceived scams. When users detected that a company had manipulated prices prior to the sale, they generated content accusing the brand of fraud.
The Consequence of Perceived Deception
The study highlighted a chain of events triggered by deceptive marketing. When users identified a fake offer, they responded with “Insults and Noise.” This category represented anger and frustration directed at the brands. The analysis showed that this negative content directly harmed the online reputation of the companies involved.
The researchers found that customer support efforts on Twitter were largely ineffective in these scenarios. The topic “Customer Support” received a negative sentiment score. This indicated that while brands attempted to manage complaints through the platform, these interactions rarely resolved the dissatisfaction caused by the initial perceived deception.
Comparing Company Performance
The study also broke down sentiment by specific companies. Brands such as Amazon and Game received largely positive sentiment scores. The analysis suggested this was because these companies offered personalized discounts and fostered engagement that users felt was authentic. However, Amazon also faced some negative noise due to an employee strike that occurred simultaneously, showing how external events infiltrate sales-focused discussions.
Conversely, retailers like Media Markt and El Corte Inglés saw higher levels of negative sentiment. The textual analysis linked this negativity directly to the “Fraud” topic. Users frequently cited these companies when discussing price manipulation strategies. The data indicated that these marketing tactics resulted in a measurable decline in brand perception during the event.
Neutral Ground
Certain topics failed to generate strong emotional responses in either direction. The broad category of “Offers and Discounts” received a neutral rating. This suggests that the mere presence of a sale is not enough to excite consumers. Similarly, promotions related to “Computers Accessories” did not generate the same enthusiasm as smartphone deals.
Scope and Boundaries
This study operated within specific parameters. It focused exclusively on the Spanish market and the technology sector during a single year. The findings reflect the behavior of Twitter users, who may not represent the entire consumer population. Additionally, the study relied on automated algorithms to interpret human language, a process verified by statistical checks but subject to the nuances of irony and slang.
Business Implications
The findings offer actionable insights for companies planning future sales events. The data indicates that standard discounts are no longer sufficient to generate positive engagement. Users respond best to “flash sales” or exclusive promotions that feel urgent and real.
The research also provides a warning regarding pricing strategies. The practice of inflating prices before a sale creates a direct path to negative reputational damage. The study shows that users actively monitor prices and share their findings. Once a company is linked to “Fraud” topics, the negative sentiment spreads quickly and is difficult to reverse through standard customer support channels.
Future Directions
The identification of these seven topics opens new avenues for academic inquiry. The researchers suggest that these themes could serve as variables in future statistical models. This would allow analysts to measure the precise impact of “Fraud” accusations on sales volume.
Future studies could also expand the geographical scope. Researchers could apply this methodology to compare consumer reactions in Spain with those in other European countries or the United States. Such cross-cultural analysis would reveal whether the sensitivity to price manipulation is a global phenomenon or specific to certain markets. The methodology established here provides a template for monitoring the pulse of digital consumerism in real-time.
