A Strategic model for service-oriented enterprises based on online reviews: the research of budget hotel chains in China

Abstract

In a complex competitive market environment, service enterprises should focus on the dynamics of consumer needs, integrate resources, and adjust their strategies to adapt to market changes. SWOT analysis is a tool that supports strategic planning and decision-making, it relies on the insight and analysis of managers. Therefore, it suffers from subjectivity and a lack of reliable inputs from the customers’ perspective. In addition, Strengths, weaknesses, opportunities and threats (SWOT) analysis cannot track the factors affecting enterprise development in real time and quantitatively determine the degree of influence of these factors. Therefore, managers are unable to make timely strategic adjustments. Online reviews that reflect consumer needs provide a new basis for dynamically formulating strategies to compensate for the shortcomings of traditional SWOT model. Therefore, this study proposes a SWOT model based on online reviews, which takes feature extraction, revised importance-performance analysis (IPA), and SWOT analysis as the framework. Through text mining and sentiment analysis methods, we construct satisfaction and importance indicators, and determine the priority of feature improvement based on opportunity algorithms, the framework can also perform dynamic SWOT analysis. The case studies of the 7Days Inn and Home Inn show that the analysis model can integrate consumer opinions, conduct a fine-grained quantitative analysis of features, and provide a basis for dynamic adjustments to corporate strategies based on horizontal and vertical comparisons between enterprises. The model is realistic and operational.

Appendices

Appendix A

Based on the factors mentioned by Liu et al. [32] and Abrahams et al. [34], we screen the factors involved in building a usefulness model. The factors finally determined include four major categories: linguistic factors, information quality factors, metadata factors and emotional factors; the specific contents are shown in Table

Table 16 Factors affecting the usefulness of online reviews

16.

Based on the 15 factors in the four categories listed in Table 16, a screening model for useful text is constructed:

$$y\left(helpfulness\right)=\alpha +\beta {L}_{f}+\gamma {IQ}_{f}+\delta {M}_{f}+\theta {E}_{f}+\varepsilon$$

(7)

where, y (helpfulness) is a categorical variable that indicates whether a text is useful. L_f, IQ_f, M_f and E_f represent linguistic factors, information quality factors, metadata factors and emotional factors, respectively.

Based on the initially established screening model, manually annotated online reviews are trained, and logistic regression analysis is performed on the training samples. Finally, the significant features are selected to determine the specific screening model for the subsequent screening of online reviews. The steps are as follows.

First, one-tenth of all the reviews are randomly selected as training and testing samples. Each selected review are marked as 0 or 1, with 0 indicating that the review is useless and 1 indicating that it is useful. Two graduate students familiar with this field manually annotate the selected reviews. If the annotation results of the two students are consistent, the annotation is considered valid. If ambiguity existed, the two students discuss it and decide together. Cohen's kappa is equal to 0.695, proving that the consistency test is passed. Second, the annotated text is divided into a training set and a test set at a ratio of 7:3. The training set is trained based on the 15 factors listed in Table 16, and the training results are subjected to logistic regression analysis while the multicollinearity problem between features is considered. Features with VIF less than 10 and significance are selected for modeling. As shown in Table

Table 17 The results of model factors

17, seven factors are eventually significant; therefore, the following prediction model is established based on the seven significant factors:

$$\begin{gathered} y\left( {{\text{helpfulness}}} \right) = \alpha + \beta_{1} L_{f} NW + \beta_{2} L_{f} ALS + \beta_{3} L_{f} NADJ + \gamma_{1} IQ_{f} NPF \hfill \\ \quad \quad \quad \quad \quad \quad \quad \quad + \gamma_{2} IQ_{f} NSPF + \theta_{1} E_{f} NOP + \theta_{2} E_{f} NON + \varepsilon \hfill \\ \end{gathered}$$

(8Ia)

A test set is used to verify the effectiveness of the model. The model evaluation indices used are: accuracy, recall, and F-score. Simultaneously, to verify the robustness of the model and determine the best classification model, we use SVM, LR, random forest, and Gaussian Bayes to verify the prediction model. The results are shown in Table

Table 18 Evaluation results of various classification models

18. It can be observed that the F-score of each classification algorithm reach a good level, and the SVM has the best performance. Therefore, we choose the SVM model to classify all reviews.

Appendix B

Considering that the online reviews in this study are relatively short and the evaluation of influencing factors is relatively concentrated, the sliding window should not be too large. Therefore, the specific identification process of the features, sentiment words and adverbs of the online comments in this study is as follows: the features mentioned in the online reviews are identified based on the constructed feature lexicon. When the corresponding feature word is identified, four windows (σ = 4) slide left and right to identify the adjectives (sentiment words) centered on the feature word. To further identify adverbs on the basis of adjectives, adverbs are identified by sliding two windows (σ = 2) left and right with the adjective as the center, thereby calculating the sentiment value. Note that the positions of the identified features and adjectives are not completely consistent, and specific procedures need to be adjusted according to the positions of the features and adjectives.

Appendix C

Strength (S): the features of the target enterprise and competing enterprise are all strengths. This part of features represent the strengths of the target enterprise. The strategy of the target enterprise is to maintain the performance of features and ensure that they are not transformed into corporate threats.

Weakness (W): the features of the target enterprise and competing enterprise are weaknesses. This part of features represents the weaknesses of the target enterprise. A basic view of the SWOT analysis is that avoiding weaknesses may ignore innovative opportunities. Thus, the target enterprise should improve the performance of these features and convert them into strengths or innovation opportunities.

Opportunity (O): the features are the strengths of the target enterprise and the weaknesses of the competing enterprise. This part of features represent the opportunities of the target enterprise, which means that the target enterprise perform better than the competing enterprise and obtains relative competitive strength. The target enterprise’s strategy is to maintain the development of these features to maintain their competitive advantages.

Threat (T): the features are the weaknesses of the target enterprise and the strengths of the competing enterprise. This part of features represents the threat of the target enterprise. This means that the target enterprise’s performance is lower than that of the competing enterprise, and the features become a relative disadvantage for the target enterprise. Target enterprise need to pay attention to these features and immediately improve them to prevent loss of potential benefits.

Appendix D

Questionnaire survey on the degree of agreement with the SWOT analysis results of 7Days Inn.

7Days Inn SWOT analysis	The degree of agreement
	4	3	2	1
Service is a strength of 7Days Inn
Location (such as near bus stops and scenic spots) is a strength of 7Days Inn
Price is a strength of 7Days Inn
Cleanliness is a strength of 7Days Inn
The problem of poorly equipped bedding products (such as bed sheets and mattresses) is a weakness of 7Days Inn
The problem of network (such as network equipment and network speed) is a weakness of 7Days Inn
The problem of surrounding environment and internal sound insulation are threats of 7Days Inn
The problem of imperfectly equipped infrastructure is a threat of 7Days Inn
Declining corporate brand recognition is a threat of 7Days Inn
The problem of inadequately equipped room facilities (such as room type and TV) is a threat of 7Days Inn
Most stores do not have catering is a threat of 7Days Inn