Keywords

1 Introduction and Theoretical Framework

The web is a mass communication medium that makes digital social interaction easily accessible and available to a global community of users. Particularly, when it is used to generate content based on people’s experience in acquiring and consuming a product or service, it becomes highly relevant as it produces information called “electronic Word of Mouth” (eWOM) referring to Online Consumer Reviews (OCR) [1]. Its importance is such that this source is recognized as a direct influence on the sales of any product [2].

One of the important places established for collecting those reviews and for eWOM communication are consumer opinion platforms [3, 4]. However, there is another resource that influences the decision-making of a potential consumer: the recommendations of opinion leaders [5].

From the consumer’s point of view, with the significant increase in the availability and variety of reviews on the web, it becomes a challenging task for users to find useful reviews that can help them decide among the restaurants of their interest. One of the main reasons for this is that web content on this topic tends to be extensive, scattered, and decentralized, which leaves users interested in certain information unsure of where to look.

On the other hand, restaurants and companies in the restaurant sector have a valuable asset in the form of reviews and opinions on social networks and rating platforms. However, it is underutilized because investing in its management and interpretation is an expense that few are willing to pay because it is not considered a priority for the business; furthermore, there is little understanding of the direct impact that the analysis of the information would generate. In short, a valuable area of opportunity in the restaurant sector that resides in the enormous amount of data, is being wasted, particularly in the consumer’s opinion.

The use of technological tools within McKinsey’s decision-making model—the consumer decision journey—is relevant and paramount for the times we live in. To understand and truly measure consumer behavior in their purchase decision, it is no longer sufficient to only consider the performance of activities that generate direct sales, as such evaluations are made in isolation from the commercial context [6]. This is where the user experience in instances with a high flow of information, as occurs in digital platforms, becomes particularly relevant.

Social media platforms with their followers also influence the decision-making process of people through the opinions and reviews influencers make; this phenomenon is known as “social media influencer marketing” [7]. This type of relatively new marketing has had huge impact, as statistic study results found: 49% of consumers rely on the recommendation of an influencer to make a purchase; if consumers feel confident in an influencer’s opinion, they are more likely to make a purchase, in this approach, brand managers need to make a smart decision when selecting the influencers to be engaged with the brand to gain meaningful trust among their customers and better sell their products or services, as findings suggest that online marketing strategies should be based on long-term relationships with influencers in order to achieve the said trust [8, 9].

Figure 1 shows a influencer marketing report from 2019 to 2023 in the United States, it is relevant as it shows a great insight on how much money it is being invested and how fast it increases, being almost $1 billion per year. Growth has also been higher than what was previously anticipated, driven by higher spending by companies already using influencer marketing, as well as new marketers leaning into it [10].

Fig. 1
A vertical bar graph for U S influencer marketing spend in billions versus years from 2019 to 2023 plots the data as follows. 2019 at 2.42 dollars, 2020 at 2.90 dollars, 2021 at 3.90 dollars, 2022 at 4.99 dollars, and 2023 at 6.16 dollars.

United States influencer marketing speed

Full size image

In the matter of restaurants, they can extensively benefit from influencer partnerships as they can provide exclusive invitations and pleasurable experiences, consequently, influencers will express satisfactory opinions in their social media on aspects of the restaurant such as food, service, location, comfort of the place, prices, style, etc. When a restaurant initiates a marketing strategy with an influencer, it would quickly increase the number of followers on the restaurant’s page on the social network and the number of bookings [11]. In this sense, it can be assumed that the number of restaurant reviews on social media is not only large but is also increasing, then it results in a great opportunity to collect them for their analysis.

This research focuses on centralizing restaurant reviews, analyzing this set of opinions and recommendations to obtain information that can be beneficial to the consumer when deciding about where to go, however, the seller also benefits if it improves the negative reviewed aspects.

2 Methodological Considerations and Results

Twitter and Instagram social media platforms were considered in this research, focusing on accounts with one thousand followers or more. The criteria considered were if influencers were reviewing specific aspects about the restaurants such as: the food, place or service, in other words, if influencers were expressing an opinion about aspects of any restaurant.

This project adopts qualitative research with a descriptive exploratory approach for which, as a first step, the identification of the candidate tools to be used as a source for the collection of reviews was carried out, in this case the social networks Twitter and Instagram and the rating platforms TripAdvisor, Yelp and Google Reviews were proposed. The coincidence criteria by which these tools were chosen is that, even though the main purpose of each of them is different, in all of them it is possible to find a significant number of restaurant reviews, which is the target information. The following Table 1 shows the characteristics according to each source.

Table 1 Relevant characteristics of the proposed information sources
Full size table

The collected reviews were strictly written in Spanish since the cohort of restaurants were from Mexico City. Once the reviews had been compiled in one same place, each one was analyzed. The elaboration of this task was completely manual, since analyzing more than 10,000 reviews would have been costly in terms of time and effort, a web tool was developed. The tool facilitated the analysis and extraction of information from each review; the criteria defined to quantify the amount of information obtained was that the number of “aspect—opinion” pairs identified by each review should be at least one; understanding as “aspect—opinion” pair to an adjective (opinion) that describes an entity (aspect). For example, in the sentence “The food is delicious”, the entity or aspect is “food”, and the adjective or opinion is “delicious”. Figure 1 shows the mentioned web tool with an example of its operation (Fig. 2).

Fig. 2
A screenshot presents 2 sections. The left section has small tabs arranged in 6 rows. The tabs have text in a foreign language. The right section has tabs arranged in 5 rows. The text on the tabs are in a foreign language.

The analysis web tool in use, showing examples of “aspect—opinion” pairs

Full size image

On the left, one can see the retrieved review to be analyzed, separated into tokens for an easier comprehension. On the right, the identified pairs “aspect—opinion” are shown as the review analyst selects them from the left section. Each pair “aspect—opinion” has a positive or negative interpretation, for example “decoration incredible” would be categorized as positive.

According to this, it was observed that not enough information was collected from the proposed social networks, Twitter and Instagram. For instance, for Twitter it was observed that most of the reviews were negative. This is attributed to the following phenomenon: Twitter is known for being a space where people can express themselves freely, which leads to a higher probability of finding hate speech and negativity in what people can say.

On the other hand, on Instagram there are several accounts intended to capture their experience through multimedia content such as photos with caption, which is where the review is written. However, in contrast to Twitter, most of them turn out to be positive. Instagram is a space in which you can easily show, and even pretend, a certain lifestyle or status, so it lends itself to showing only those pleasant, joyful and desirable experiences.

Given the circumstances, the decision was made not to include reviews from social media as part of the information collected. In relation to the evaluation platforms Yelp, TripAdvisor and Google Reviews, the reviews were more objective. It can be explained by understanding that the objective of the evaluation platforms is to transparently communicate whether they recommend a restaurant or not, through the narration of their experience and the presentation of the aspects of the restaurant that stand out, whether due to satisfaction or dissatisfaction of the product or service.

The case study analysis exercise began in January 2022, the review collection began in August 2022 and ended in June 2023, using automated processes to extract reviews using web scrapping, in addition to APIs provided by the rating platforms themselves. They were collected in text format, and was the same for both extracting forms, thus consolidating a single repository with these reviews.

3 Discussion

This research highlights the use of different information sources to gather diverse restaurant reviews, relevant for the acquirement of surfaced key observations about how the subsequent review analysis process enabled a better comprehension about the quality of the retrieved information in accordance with the purpose of delivering a good information repository. Consequently, social media reviews were omitted in favor of more objective reviews, predominant on rating platforms.

In addition, the adoption of technological tools has become pivotal, as said review analysis would not have been possible in the same way without the developed web tool investing the same resources. Furthermore, its incorporation into the understanding and measurement of the consumer behavior when he makes purchase decisions should now be mandatory, since valuable information can be obtained that at first sight and in a singular way, is not clearly perceptible until it is analyzed in a set.

4 Conclusions

In conclusion, this project manages to address this area of opportunity by centralizing both recommendations from opinion leaders (influencers) on social networks and consumer reviews on rating platforms, presenting them in a way that closely approximates the objective reality for a user regarding the aspects of a restaurant. Nevertheless, it also provides relevant information to the restaurant sector, therefore two fields are expected to be benefited, consumers and restaurants; while diners can improve the time and the way they decide where to eat based on the set of reviews by not having to look for information on each existent platform, restaurants can evaluate aspects of their own business such as consumer consumption habits, their behavior in relation to various factors as it could be weather conditions, and most importantly, improvement areas based on what is reviewed and the opinions of the consumers.

Finally, the analysis of this information leads to the necessity of a software tool that implements a Natural Language Processing—Artificial Intelligence (AI) subarea—learning model. The information collected in this project would be effective for training the AI model, therefore, the analysis of the reviews could be done in a faster and automatic way; thus, a new global review could be generated based on all the reviews of each restaurant, facilitating the customer’s decision-making process.