From a servicescape perspective, the hotel indoor environment is not merely a physical background. [6] show that spatial layout can shape value perception and customer loyalty in theme hotels, and [14] describe the substantive servicescape as a source of customer inspiration. Based on this view, the present study defines the hotel indoor environment as a set of cues that can influence emotion, perceived value, satisfaction and behavioral intention.

Guest assessment therefore begins with direct perception. [9] demonstrate that illuminance, color and decoration style affect visual comfort in hotel lobbies; [8] use guest experience data to identify innovation opportunities in hotel interior design; and [11] show that artwork and environmental awareness can be incorporated into urban hotel interiors. These studies justify examining whether colors are coordinated, materials feel refined, lighting is comfortable, the room is quiet, temperature and ventilation are appropriate, and layout supports use.

IEQ is a central basis for comfort and satisfaction. [23] use text mining to assess IEQ in Chinese budget hotels, while [17] identify IEQ complaints and dissatisfaction from Booking.com reviews. [21] show that hotel-room VOC contamination can create air-quality and health-risk concerns. These studies explain why this paper separates light, thermal conditions, air quality, acoustics and visual environment into distinct perception channels.

The emphasis on bodily comfort is also supported by [31], who connect IEQ with tourist accommodation ratings, and by [12], who relate IEQ to hotel guest behaviors. Therefore, poor lighting, inadequate air conditioning or ventilation, weak sound insulation, odor and stale air are treated not as minor operational complaints but as environmental signals that can influence perceived quality, hygiene inference and management evaluation.

Online reviews provide an important basis for identifying experience elements without imposing researcher assumptions too early. [3] demonstrate how a hotel dictionary can be created for sentiment analysis using TripAdvisor reviews; [20] apply topic modeling to hotel service experience; and [26] compare BERT and ERNIE for sentiment analysis of Chinese hotel reviews. These studies support the use of review texts as analyzable evidence of consumer experience.

The specific review-mining logic also follows broader hospitality text-analysis work. [4] review text analysis methods in tourism and hospitality, [22] show that review valence and content jointly affect review value, and [27] demonstrate that topic and sentiment information can predict hotel demand. Accordingly, this study uses review data to support questionnaire construction and to interpret hotel guests’ natural expressions of indoor environmental experience.

Interior design style refers to the integrated impression of perceptible color, materials, furniture, furnishings, spatial order and atmosphere. [15] describe the new Chinese style as a combination of traditional cultural elements and contemporary interior design development, while [25] discuss hotel B&B interiors through spatial narrative. [16] links indoor environmental art decoration with visual comfort evaluation. These studies support the definition of style as a perceptual system constituted by color, material, form, lighting, display, scale and spatial order.

Non-lighting visual elements in this study therefore refer to visual design factors other than daylight and artificial lighting, including color, materials, furniture, soft furnishings and spatial layout. [7] further connect sustainable interior design criteria with hotel operational effectiveness, which suggests that visual and spatial components should be considered together with environmental comfort rather than as isolated aesthetic surfaces.

Interior design style may influence first impressions and aesthetic preference, but its effect in real consumption settings may be diluted or moderated by comfort-related experiences. [18] show that room aesthetics can affect booking preference, yet [31] and [1] indicate that IEQ and comfort also matter for accommodation evaluation. Therefore, a distinctive room style may still receive a low evaluation if it is noisy, poorly ventilated or thermally uncomfortable.

This distinction is also why the hypotheses test style-group differences rather than assuming that style alone determines evaluation. [29] provide direct evidence that decoration styles can influence subjective hotel-environment evaluation, whereas [6] and [14] suggest that spatial layout and servicescape cues work through perceived value, loyalty or inspiration. The present paper therefore treats style as a cue embedded in a complex consumption situation rather than as a stand-alone determinant.

Because different interior design styles vary in color, material, spatial layout, furniture, decorative details and lighting atmosphere, they may be related to guests’ indoor environmental perception. This expectation follows [29], who compare decoration styles in hotel IEQ evaluation, and [9], who identify visual comfort effects of light, wall color and decoration style. On this basis, the following hypotheses were tested:

H1: Interior design style groups differ significantly in their evaluation of non-lighting visual elements.

H2: Interior design style groups differ significantly in their evaluation of the light environment.

H3: Interior design style groups differ significantly in their evaluation of the thermal environment.

H4: Interior design style groups differ significantly in their evaluation of air quality.

H5: Interior design style groups differ significantly in their evaluation of the acoustic environment.

H6: Interior design style groups differ significantly in their overall environmental evaluation.

A mixed-method design was used: online review text analysis, questionnaire measurement and statistical testing. The design follows the logic of [3] and [20], who use review texts to extract hotel-experience information, and it is consistent with [4], who summarize text analysis as an important approach in tourism and hospitality research. The logic was to identify guests’ real concerns in online reviews, combine those findings with prior literature to construct a questionnaire, and then use questionnaire data to test style-group differences and predictors of overall environmental evaluation.

In practical terms, the review-mining stage followed prior hotel text-analysis studies by using naturally occurring online comments to identify guest-experience themes. [26] and [27] show that sentiment and topic information in hotel reviews can reveal meaningful consumer responses, so the present questionnaire stage translated those themes into measurable IEQ and servicescape constructs rather than treating review words as isolated frequencies.

The online review data were collected from hotel reviews on the Ctrip App by using Octopus Collector. Similar to [17] and [31], who use platform-based accommodation reviews to study IEQ and guest ratings, the present corpus was designed to identify recurring hotel-environment themes rather than to estimate platform-wide prevalence. The review-extraction window was fixed before collection as one month, from 24 April 2026 at 13:21 to 24 May 2026 at 13:21. This time stamp records the actual start and end of web data extraction, and the corpus is described as a time-bounded, platform-based purposive sample rather than a probability sample of all Ctrip hotels. The sampling frame was defined as Ctrip hotel pages that met four criteria: 1) the hotel page contained identifiable indoor-space photographs or descriptions that allowed the room/interior style to be recognized; 2) the page provided public post-stay user reviews rather than promotional descriptions; 3) the hotel was located in Chinese Mainland urban tourism or business destinations and represented commonly encountered commercial hotel settings; and 4) the review text contained substantive comments on room, environment, facilities, service or stay experience. Search and filtering focused on hotels that could be grouped into recognizable interior-style categories and excluded homestays, purely scenic-area inns, apartments without hotel-like service, pages with too few substantive comments, duplicate texts, advertisements and comments unrelated to hotel stays. The retained texts were public post-stay comments visible on the selected Ctrip review pages during the one-month extraction window. A total of 1298 review texts were collected. After cleaning, 1267 valid reviews remained, producing a validity rate of 97.61%. The data were used only for academic analysis and did not include identifiable personal information.

Cleaning included format standardization, duplicate removal, exclusion of empty or non-substantive comments, deletion of garbled text, advertising content, emojis, irrelevant symbols and content unrelated to hotel stays. Following [3], a user dictionary was constructed around space and vision, lighting, thermal environment, air quality, acoustics, facilities, hygiene, service and amenities. The cleaned texts were imported into KH Coder for high-frequency word statistics and co-occurrence semantic network analysis. Consistent with [22], service, hygiene and value-related words were retained as contextual evidence of integrated hotel experience, but they were not modeled as dependent environmental-perception dimensions because the theoretical focus of the survey was indoor environmental quality and interior spatial experience. The online review data and text-processing procedures are summarized in Table 1.

The questionnaire was developed from the text analysis results and IEQ-related literature through a review-mining-to-item procedure. First, high-frequency words and co-occurrence clusters were screened to identify recurrent experience themes, including room, environment, facilities, hygiene, service, breakfast, sound insulation, air conditioning, smell, light, decoration and value. Second, themes directly related to indoor environmental perception were mapped to five environmental channels. This mapping follows [23], [17] and [21], who respectively highlight hotel IEQ complaints, online-review-based IEQ dissatisfaction and air-quality/VOC concerns. Third, prior servicescape, IEQ and multisensory hotel-experience literature was used to convert these themes into measurable Likert items and to add an overall environmental-evaluation dimension.

The item-mapping procedure was therefore not arbitrary. Review terms such as sound insulation, air conditioning, smell, light and decoration were connected to IEQ and servicescape constructs already discussed by [1], [12], [24], [6] and [19].

Table 1. Online review data and text processing summary.

At the same time, themes such as hygiene, service and value were excluded from the final structural model because they represent service-operation or value-evaluation constructs rather than indoor environmental perception itself; they were retained only to interpret the broader consumption context.

Interior design style groups were defined at the first survey stage as respondents’ recognition of the main style of the hotel they had recently stayed in. The questionnaire provided concrete examples to reduce ambiguity: modern minimalist style, Nordic style, Japanese wabi-sabi/modern Japanese style, New Chinese style, American/British pastoral style, light luxury style, industrial style and other. For statistical analysis, conceptually adjacent categories with small cell sizes were combined into four style groups: modern/minimalist style (n = 73), Nordic/Japanese style (n = 41), New Chinese/American pastoral style (n = 45), and light luxury/industrial/other style (n = 44). Case assignment was therefore based on the respondent’s reported recent-stay hotel style, with adjacent small categories merged to preserve interpretable groups and adequate sample size for ANOVA/Welch comparison.

The questionnaire was distributed online through Wenjuanxing. The survey was launched on 24 April 2026, the same date as the start of the Ctrip review-extraction window, so that the review-mining and questionnaire stages referred to the same research period. Respondents were consumers with actual hotel stay experiences in the previous 12 months and who could clearly recall the room style and environmental details of a recent stay. The survey was anonymous, and all data were used only for academic purposes. In total, 204 questionnaires were collected. One questionnaire with all scale items unanswered was removed, leaving 203 valid cases for data checking and main analysis.

A univariate missing-data check across 31 scale items showed that 30 core items had very low item-level missing rates (approximately 0.5%, equivalent to one missing response per affected item), while one reverse-coded item had a missing rate of 13.3% (27/203). This higher missing rate was most likely caused by the negative wording of the item, “The room design style was messy and the visual experience was poor,” which differed from the positively worded items and may have increased respondent hesitation or skipping. Because the missing rate was below 15% and the item was theoretically useful for response-quality checking, the item was retained and reverse-coded before scale construction. To avoid confusion between missing-data treatment and the inferential models, the revised analysis distinguishes three data sets: complete cases (n = 176) for reliability and exploratory factor analysis, available composite-score cases (n = 203) for the main ANOVA/Welch and regression tables, and five multiply imputed data sets used only as sensitivity checks. Composite dimension scores were calculated when respondents had valid answers for the majority of items in the relevant dimension, so a missing response to the reverse-coded item did not automatically remove the respondent from the main composite-score analyses.

The analysis proceeded through reliability testing, KMO and Bartlett tests, principal component analysis with varimax rotation, descriptive statistics, one-way ANOVA/Welch tests with Bonferroni correction and hierarchical regression. When Levene’s test indicated unequal variances, Welch’s robust test was interpreted. Because six environmental dimensions were tested simultaneously, the Bonferroni-adjusted significance threshold was 0.05/6 = 0.0083. For regression, overall environmental evaluation was the dependent variable, and non-lighting visual elements, light environment, thermal environment, air quality and acoustic environment were predictors. Multiple imputation in SPSS 26 used five imputed data sets, a linear-regression imputation model and a fixed random seed of 999; however, the main inferential tables report the non-imputed composite-score results to keep the reported F values, degrees of freedom and VIF values directly reproducible. A sensitivity comparison between the complete-case results and imputed-data checks did not change the substantive conclusions: style-group differences remained nonsignificant, and thermal and acoustic environment remained the most stable positive predictors of overall environmental evaluation.

The high-frequency word analysis showed that guests’ core comments focused on service process, room space, hygiene, facilities, amenities and value judgment. This pattern is consistent with [33], who derive customer preferences from aspect-level sentiment analysis of hotel reviews, and with [13], who mine large-scale online hotel reviews to understand satisfaction patterns. High-frequency words included service, hotel, room, breakfast, environment, hygiene, facilities, front desk and value for money. These results indicate that guests did not evaluate hotels only through interior design style; instead, they integrated spatial experience, cleanliness, facility conditions, service quality, supporting amenities and perceived value into their stay judgments.

The result also matches [30], who uses LDA topic modelling to analyze guests’ aesthetic experience with lifestyle hotels, and [27], who show that review topics and sentiment have predictive value. Thus, the Ctrip review findings are not merely descriptive word counts; they help explain why the later questionnaire had to include multisensory environmental channels rather than a single style-preference item.

The co-occurrence semantic network further showed that room, hotel, service, environment, hygiene and breakfast formed a prominent core network. In line with [20], who use topic modeling to identify hotel-service experience themes, this network indicates that guests tended to evaluate indoor environment, cleanliness, service experience and amenities as a whole rather than as separate attributes. These findings support the construction of a multidimensional questionnaire and confirm that hotel indoor environment evaluation should be treated as an integrated consumer experience rather than as a purely visual or stylistic judgment.

The complete-case reliability and construct-validity tests included 176 original cases. This number is consistent with the missing-data check because 27 respondents had at least one missing scale response, leaving 176 respondents with complete data across all 31 scale items. The overall scale contained 31 items and showed good internal consistency, with Cronbach’s alpha = 0.861 and standardized alpha = 0.860. Each subscale also exceeded 0.80. The KMO value was 0.802, and Bartlett’s test of sphericity was significant, indicating that the data were suitable for dimensional reduction.

Exploratory factor analysis extracted six common factors with eigenvalues greater than 1. The cumulative variance explained was 59.298%, which was acceptable for this exploratory study. The six factors were interpreted as non-lighting visual elements, acoustic environment, light environment, thermal environment, overall environmental evaluation and air quality. The structure matched the questionnaire design and was therefore used in subsequent analyses. The reliability and construct validity results are summarized in Table 2. Appendix A reports the 31 questionnaire items, their assigned dimensions and primary factor loadings so that readers can evaluate how the six dimensions were operationalized.

Table 2. Reliability and construct validity summary.

Note. Reliability and exploratory factor analysis were calculated using original complete cases (n = 176). Main ANOVA/Welch and regression analyses used composite-score cases (n = 203); multiple imputation was used only for sensitivity checking.

The main analysis sample consisted of 203 respondents. Women accounted for 51.7% and men for 48.3%. Respondents aged 19 - 25 formed the largest age group (35.0%), followed by those aged 26 - 35 (25.6%) and 36 - 45 (22.2%). Educational backgrounds and occupations were diverse, although teachers and enterprise/company employees formed sizable groups. In terms of monthly disposable income, 43.3% reported 3000 yuan or below, and 35.0% reported 3001 - 6000 yuan. The demographic characteristics of the respondents are presented in Table 3.

Table 3. Demographic characteristics of respondents.

To test whether interior design style groups differed in perceived environmental evaluation, interior design style was used as the grouping variable, while the six dimension means served as dependent variables. The four style groups were modern/minimalist style (n = 73), Nordic/Japanese style (n = 41), New Chinese/American pastoral style (n = 45), and light luxury/industrial/other style (n = 44), totaling 203 valid composite-score cases. Levene tests were first used to assess homogeneity of variance. For non-lighting visual elements, light environment, thermal environment, air quality and acoustic environment, Levene p values were below 0.05, so Welch robust tests were interpreted. Overall environmental evaluation met the homogeneity assumption, so the ordinary ANOVA result was interpreted.

The tests showed no significant differences for non-lighting visual elements, light environment, thermal environment, air quality, acoustic environment or overall environmental evaluation. Although the acoustic environment result approached 0.05, it did not reach statistical significance and was clearly nonsignificant after Bonferroni correction. The ANOVA and Welch test results by interior design style group are presented in Table 4. Therefore, none of the six hypotheses was supported.

A hierarchical regression was used to determine whether the environmental perception dimensions explained overall environmental evaluation. Model 1 included only control variables: gender, stay purpose, room type and decoration style. Model 2 added five environmental perception dimensions: non-lighting visual elements, light environment, thermal environment, air quality and acoustic environment. The reported regression used the 203 composite-score cases, while complete-case and imputed-data sensitivity checks produced the same substantive pattern. Model assumptions were acceptable. All predictors had VIF values between 1.050 and 1.487, and the Durbin-Watson statistic was 2.063. The hierarchical regression results and model-fit statistics are presented in Table 5.

Table 4. ANOVA and welch tests by interior design style group.

Note. Bonferroni correction was applied across six tests; adjusted threshold = 0.0083.

Table 5. Hierarchical regression results for overall environmental evaluation.

Note. Dependent variable = mean overall environmental evaluation. Model 2 F is the overall model F based on R squared = 0.313, N = 203 and 12 predictors; Delta F reports the block-change test after adding the five environmental perception predictors. **p < 0.01.

Table 6. Regression coefficients for environmental perception dimensions.

The regression coefficients for the five environmental perception dimensions are presented in Table 6. Model 1 had weak explanatory power (R squared = 0.034, adjusted R squared = 0.012, F = 0.983, p = 0.445). After the environmental dimensions were added, Model 2 improved substantially (R squared = 0.313, adjusted R squared = 0.277, overall F = 7.214, p < 0.001). The environmental-perception block also produced a significant model improvement (Delta F = 15.428, p < 0.001). In Model 2, thermal environment evaluation had a significant positive coefficient (B = 0.221, p = 0.004), and acoustic environment evaluation also had a significant positive coefficient (B = 0.234, p = 0.001). Non-lighting visual elements were marginally related to overall evaluation (B = 0.144, standardized beta = 0.136, p = 0.050). Light environment and air quality were not significant in this sample.

The study contributes to hotel management and consumer behavior research by shifting the empirical context from controlled experimental comparison to real consumption settings. [28] and [29] provide important controlled or comparative design evidence, but the present results show that style groups did not produce robust differences in non-lighting visual elements, lighting, thermal environment, air quality, acoustics or overall environmental evaluation after Bonferroni correction.

This interpretation is consistent with [29], who compared decoration-style effects on subjective hotel-environment evaluation, but the present study extends that logic by testing whether such style differences remain robust in a real consumption sample. The nonsignificant style-group findings also align with IEQ studies suggesting that concrete comfort factors, especially thermal and acoustic conditions, may be more decisive than broad style labels in post-stay evaluation.

The findings do not mean that design style is unimportant. Instead, they support the interpretation of style as a cue that operates through specific experience channels. This interpretation is consistent with [18], who connect room aesthetics with guest booking preference, and with [9], who show that visual factors can affect comfort. However, the stronger coefficients for thermal and acoustic environments indicate that bodily comfort channels may be more decisive than style labels in this data set.

In other words, the results refine rather than reject the design-style literature. [28], [18], [11] and [16] support the experiential value of color, aesthetics, artwork and environmental art, while [1], [12], [23] and [31] show that comfort-based IEQ factors are strongly connected with satisfaction, behavior, complaints and ratings. The present results suggest that these two lines of evidence should be integrated.

The online review results support a multisensory understanding of hotel experience. [24] explicitly conceptualizes the hotel stay as a multisensory experience, while [8] use guest interior-design data to identify innovation opportunities in urban hotels. The frequent Ctrip references to room, environment, hygiene, facilities, service, breakfast and value for money therefore show that aesthetic experience is embedded within broader stay experience.

Theoretically, the study integrates interior design style and IEQ within one framework of guest environmental perception. [1], [12] and [32] show why IEQ matters for comfort, behavior and satisfaction, while [15], [25] and [16] explain interior style, spatial narrative and visual comfort from a design perspective. Combining these streams allows the study to move beyond stylistic preference and to show that style operates within visual, thermal, acoustic, air-quality and overall experience dimensions.

Methodologically, the study demonstrates a review-mining-to-questionnaire pathway. It adapts the text-analysis logic of [3], [4], [26] and [13] by using online comments to identify natural post-stay expressions, then uses survey data to test dimensions and predictors. This mixed approach reduces the risk of relying exclusively on researcher-defined dimensions and improves contextual fit for hotel consumer behavior research.

Compared with studies that rely only on text mining or only on questionnaires, this design uses online reviews to generate context-sensitive dimensions and then uses survey data to test their relationships statistically. This is consistent with the methodological direction suggested by [4], [3], [13] and [26], but it adds a design-management focus by linking review themes to indoor environmental perception scales.

Practically, hotels should not rely solely on visually distinctive style labels when updating interior spaces. The findings align with [7], who connect sustainable interior design components with hotel operational effectiveness, and with [21], who show that air-quality issues such as VOC contamination can create health-risk concerns. Hotels should therefore use review monitoring to identify recurring experience problems such as poor sound insulation, uncomfortable temperature, weak ventilation, odor and uncomfortable lighting, then translate them into design renovation and operational management actions.

This practical recommendation follows directly from prior evidence. [17] and [23] showed that review mining can identify IEQ complaints, [21] highlighted indoor air-contamination risks in hotel rooms, [32] linked IEQ to satisfaction and ratings, and [9] connected visual comfort with hotel-lobby experience. Therefore, renovation decisions should combine style positioning with measurable improvements in ventilation, odor control, lighting comfort, sound insulation and thermal stability.

Several limitations should be acknowledged. First, the questionnaire sample was collected online and was concentrated in one regional context, which limits generalizability. Future studies should expand to larger samples across different regions, hotel grades and cultural contexts. Second, subgroup sizes by interior design style may have been unbalanced. Future research should use larger and more balanced style groups to improve the stability of comparison tests. The revised sampling-frame description improves reproducibility, but the corpus should still be understood as a platform-based and purposive review sample rather than a probability sample of all hotels.

Third, the study relies primarily on subjective questionnaire responses and online reviews. Future work should integrate objective environmental measurements such as illuminance, color temperature, indoor temperature and humidity, carbon dioxide, PM2.5, VOCs, noise decibels and reverberation time. This recommendation follows [17], who show the value of review-based IEQ complaint mining, and [21], who demonstrate the importance of directly measuring VOC contamination in hotel rooms. Fourth, the online review data came from Ctrip; platform mechanisms and user expression habits may affect review content. Future research could compare multiple booking platforms to test cross-platform robustness.

Fifth, interior design style classification relied on respondents’ recognition and self-report, which may include subjective inconsistency. Future work could combine expert evaluation, computer vision and field investigation to classify style more objectively and to capture the spatial narrative depth of hotel interiors.