The application of the internet for sales and marketing is currently commonplace due to the advancement its ever sophisticated and powerful technology. With a population of 277.7 million, Indonesia has such a comparatively high rate of social media usage. According to data collected from the CNBC Indonesia website, Indonesia have 204.7 million internet users in 2022, rise from 202.6 million in 2021. The social media sites which Indonesian most frequently access are Facebook, Twitter, Instagram, Whatsapp, Instagram and Youtube. E-marketing is the marketing part of electronic commerce, which includes the work activities of a business whose purpose is to advertise, promote and sell products and services through the internet. Also, a system like this for marketing can operate continuously for a full 24 hours utilizing nothing more than a computer linked to the internet to promote all of a company's goods [1]. Social media marketing, for example, can greatly improve sales while requiring little financial investment on the part of the business.

Social media, according to Nasrullah [2], is a platform for media that emphasizes users' existence and supports their activity. Consumers purchase decisions are directly affected by social media advertising. Consumers have a customer information search phase in the purchasing, according [3]. Marketers may offer this information, or customers may look for it on social media. Social media is already frequently used by businesses with the intention of marketing things and building client relationships. Because consumers will find it hard to find information about a product being sold by these marketers without the use of social media, social media use at this time has a significant effect on improving sales. As a result, social media use is currently a basic requirement for marketers in their marketing activities.

According to [4], word-of-mouth marketing refers to consumer behavior in which they inform other consumers about brands, products and services. Word of mouth marketing is also significant while some customers like to visit the product in person to learn more about it before making a purchase decision from the marketer.

Table 1: Statistics for Citra Florist Sales in 2020–2022

Source: Citra Florist Flower Shop

It is evident from the table above that sales have fluctuated downward, which points to the issue of weak customer buying intentions. The issue of less purchasing at Citra Florist is a problem brought on by ineffective promotional strategies, like social media that doesn't function. Nowadays, social media is something that is crucial for society since it makes it simple for people to get different kinds of information. Social media on a florist's brand doesn't function well because the florist's brand mainly emphasizes forming connections and handing out brochures. In order to stay connected with customers, there is a requirement for effective social media management. However, there are issues with word-of-mouth, where these initiatives are still inactive because the florist's reputation dictates that word-of-mouth advertising only occurs on significant holidays like Mother's Day and Valentine's Day.

People communicate with their communities, groups of friends and family member’s kind of reference group on social media in order to find fascinating content to meet their needs. Information received from friends or family and the surroundings have just as much of an impact as information found in advertisements. Information from friends, neighbors, or family will reduce the risk of purchasing because these individuals are knowledgeable sources of product information. In general, reference groups play a significant role in helping consumers choose a brand or product because they provide knowledge on items that is complementary to that provided through social media. Most prospective customers will research or consult others before choosing to use a product or service, concerning a product or service in advance, like the intention to buy products for a floral bouquet. Not least through word of mouth (WOM), also referred to as obviously "word of mouth," which is the constant transmission of knowledge from one person to another. Depending on their demands, they are looking for various types of information.

According to [5], word-of-mouth advertising refers to conversations among individuals about goods and services. This exchange can take the shape of a dialogue or a testimonies. The researcher is interested in researching "The Impact of Social Media Advertising and Word of Mouth on the Decision to Buy Flower Bouquets at Citra Florist Surabaya" based on the description of the problem.

The Citra Florist Surabaya, which is situated on Jl. Embong Ploso 2B, Embong Kaliasin, Kec. Genteng, Surabaya City, East Java 60271, was the location of this research. Although there are issues with decreasing sales and a lack of customer buying enthusiasm, the location of Citra Florist Surabaya was chosen based on the fact that the location is a flower store with a main aspects online sales system. The independent variables in this study are social media advertising and word-of-mouth and the dependent variable is purchase intention. The purpose of this study was to examine how these variables were used and how they affected consumers' propensity to buy at Citra Florist Surabaya.

According to [6], the population is the total number of elements or subjects that have specific characteristics that were chosen by the researcher to be researched before conclusions are taken. Customers of Citra Florist Surabaya who had made at least one purchase there made up the sample of the study that was to be looked at. The sample is thought to be representative of the population since it is a subset of the population with many of the same characteristics [7]. A number of existing populations were used as samples for the study due to time and financial constraints. Purposive sampling was the method that was employed and it involved the researcher selecting a sample from the consumers that encountered while doing the study. Customers who are 17 years of age or older and have made purchases from Citra Florist Surabaya meet the criteria.

A data analysis method called descriptive analysis was utilized to examine the characteristics of respondents (consumers) at Citra Florist Surabaya. The Likert scale, a standard measurement seen in surveys, is used in this study's variable measurement. The Likert scale is a rating system based on how respondents responded to queries about the idea or variable under study. to assess the Likert scale used to evaluate the variables Social Media Advertising (X1), Word Of Mouth (X2) and  Purchase Decision (Y) in order to analyze respondents' replies about the variables studied. The SmartPLS 3.0 software, which is effective for studying partial least squares data from structural equation models, will be used to process the statistical data. The measurement model (outer model), which contains validity and reliability tests and the structural model (inner model), which includes R-square for the coefficient of determination and path coefficients for testing hypotheses, both are analyzed as part of the evaluation of the PLS model.

Characteristics of Citra Florist Consumers

Age, gender, type of employment and  where do consumers get information about the citra florist. The characteristics of customers who purchase flower bouquets from Citra Florist Surabaya are known according to the percentage distribution of the characteristic data from the respondents.

Outer Model Evaluation

In order to evaluate the measuring model's viability in terms of validity and reliability, the outer model is evaluated. Using convergent validity and discriminant validity ways, the degree of validity in the evaluation of the outer model with reflection indicators is sought, while the level of reliability is sought through composite reliability approaches.

Convergent Validity

Table 3's research findings display the loading factor value, which can be used to illustrate the relation between indicators and latent variables. To reflect or reflect the latent variable correctly, the loading factor value on the indicator must be at least higher than 0.7.

Table 2: Characteristics of Citra Florist Consumers

Source: Primary Data Processed (2023)

Table 3: Output Outer Loading

Source : Results of Data Processing on SmartPLS 3.0 (2023)

Analyzing the results in the outer loading table for each latent variable Social Media Advertising, Word of Mouth and Purchase Decision will be used in this study's validity testing. If the individual reflection measure has a correlation value of less than 0.7, it is considered valid [13]. Because all loading factor values for each indicator are more than 0.70, it is known from the findings of the outer loading output above that the loading factor for all indicators of each construct can be said to meet the convergent validity criteria.

Discriminant Validit

The cross loading values between the indicators and their constructs show the discriminant validity of the reflection indicators. The cross loading output of the estimation outcomes using the PLS Algorithm with SmartPLS 3.0 software is shown in the Table 4.

The correlation between each indicator and the construct is stronger than the correlation between the other constructs, as can be seen from the cross leading output findings above. The latent construct can therefore be claimed to predict its indicators more accurately than other construct indicators. Comparing the square root of the Average Variance Extracted () of each construct with the value of the latent variable correlation, or correlation between constructs, is another method of assessing discriminant validity. If the average variance extracted (AVE) square root value of each construct is higher than the latent variable correlation value or can be determined by looking at the AVE value, the model is deemed to meet the requirements of discriminant validity. If the AVE value is greater than 0.5, the model is considered to be valid. The PLS Algorithm with SmartPLS 3.0 software's AVE output and latent variable correlation estimate results are shown in Table 5 and 6.

The AVE square root value for each construct is bigger than the correlation value of each construct to other constructs, according to tables 5 and 6, which provide this information. For instance, the correlation between social media advertising and word-of-mouth is 0.326, while the correlation between social media advertising and purchase decisions is 0.558. The AVE root value for the social media advertising construct is 0.888, which is higher than both of these values. Which is why, it may be said that this model's constructs are all discriminantly valid. Thus, the latent variable may typically account for more than half (50%) of the variance of each indicator.

Composite Reliability

Construct reliability that can be analyzed using the Cronbach alpha and composite reliability metrics. If a construct has a composite reliability rating and a Cronbach's alpha of better than 0.70, it can be deemed reliable [8]. The output of the PLS Algorithm using SmartPLS 3.0 software's composite reliability and cronbach's alpha estimate results is shown in Table 7.

According to the reliability test findings in the table above, all variables are considered reliable if their composite reliability and Cronbach's alpha values are more than 0.7. This indicates that the consistency of all the study's instruments is good.

Inner Model Evaluation

Inner model testing can be done by looking for the coefficient of determination (R2), Effect Size (F2), Stone-Geisser (Q2), Goodness of Fit Index.

R-Square (Coefficient of Determination)

The R-Square value is used to determine how much influence the independent latent variable has on the dependent latent variable.

The structural model (inner model) in this study is categorized as moderate, which is more than 0.33 but less than 0.67, based on the aforementioned table. The endogenous construct of purchase decisions' R-Square interpretation is 0.498. Hence it can be claimed that the constructs of social media advertising and word of mouth influence the purchase choice construct by 49% and other variables not included in this study influence the remaining 51%.

Table 4: Output Cross Loading

Source : Results of Data Processing on SmartPLS 3.0 (2023)

Table 5: Output AVE and

Source : Results of Data Processing on SmartPLS 3.0 (2023)

Table 6: Output Latent Variable Correlation

Source : Results of Data Processing on SmartPLS 3.0 (2023)

Table 7: Output Composite Reliability dan Cronbach Alpha

Source : Results of Data Processing on SmartPLS 3.0 (2023)

Table 8: Output R-Square (R2)

Source : Results of Data Processing on SmartPLS 3.0 (2023)

Analysis Effect Size (F2)

According to Ghozali [8], changes in the value of R2 are used to determine if the measurement of exogenous latent variables on endogenous latent variables has a significant impact (F2). The criteria for F2 value are 0.02 for a minimal impact, 0.15 for a moderate impact and 0.35 for a substantial (big) impact at the structural level [8].

According to the aforementioned data, social media advertising has an F2 value of 0.297, which indicates that it has a modest or little impact on explaining the endogenous variable of purchase decisions. The impact of word of mouth on purchasing choices is currently 0.372. So, it is understood that word-of-mouth exogenous variables have a significant (high) impact on the ability of endogenous variables to explain purchasing decisions.

Analysis Stone Geisser (Q2)

The second test is called the Q2 test, which measures how well the observed values generated by the model and the parameter estimates, or how well the path model can predict the values of the original data, can be predicted. The results of the blind calculation in the construct validated redundancy section show Q2. When the Q2 value is larger than 0, the model is considered to be predictively relevant; when it is less than 0, it is not.

The Q2 value is 0.385, as can be seen from the table above. The model's predictive relevance is already present because the Q2 value is greater than zero. In other words, social media advertising and word-of-mouth factors have a good level of predictability for future purchases.

Analysis Goodness Of Fit Index (GoF)

The total model is validated using the Goodness of Fit (GoF) index. The average root value of communalities is multiplied by the average root value of R-Square to provide the findings of the GoF test. The GoF calculating formula is as follows. 

Where Com is the model's average measure of communality and R2 is the model's average measure of R2. The value of R2 in the purchase decision variable, which is equal to 0.498, can be used to get the average value of R2. In the meanwhile, model measurements utilizing the blindfolding technique in the construct validated communality section can be used to determine the communality value of each variable. The following table shows the communality score on average.

Based on table 11 above and the average value of R2, the GoF values are: 

The calculation results above yield a GoF value of 0.589, which indicates that the model has a large GoF, more than 0.36, indicating that the measurement model (outer model) and the structural model (inner model) are feasible. The higher the GoF value, the better it is at describing the research sample.

Hypothesis Test

The route map that follows is based on the PLS algorithm's findings and it shows each indication in the study model Social Media Advertising as X1, Word of Mouth as X2 and Buy Decision as Y.

The resampling bootstrapping approach was utilized in this study to test the hypothesis in order to ascertain the impact of exogenous constructions on endogenous constructs as well as the impact of endogenous constructs on endogenous constructs. In this study, the positive values of the T-Statistics, P-Values and beta coefficients (Original Sample) were used to test the hypotheses. If the T-Statistics value > T-Table is 1.645 and the P-Value significance level value is 0.05, the study hypothesis can be considered to have been accepted [8]. The following table illustrates the importance of testing the study's hypothesis.

The path coefficients table above shows the outcomes of this study's test of the hypothesis, which may be further explained as follows:

• H1: The intention to purchase flower bouquets from Citra Florist Surabaya is positively and significantly impacted by social media advertising

According to the route coefficient table, the social media advertising variable has a P-Value of 0.043 because (H1) is approved if P-Value 0.05. The link has a positive impact, as seen by the path coefficient value (Original Sample), which is positive and 0.409 in this case. The decision to buy flower bouquets from Citra Florist Surabaya can be inferred to have a positive and significant impact from social media advertising.

• H2: The intention to purchase flower bouquets from Citra Florist Surabaya is positively and significantly impacted by word of mouth

According to the route coefficient table, the word-of-mouth variable has a P-Value of 0.020 since (H2) is accepted if P-Value 0.05. The decision to buy flower bouquets from Citra Florist Surabaya is influenced positively and significantly by word of mouth, as shown by the path coefficient value (Original Sample) of 0.457, which is positive and indicates that the association has a favorable effect.

The following is a discussion of the findings from the hypothesis testing on the impact of word of mouth and social media advertising on purchasing decisions:

The Effect of Social Media Advertising on Purchasing Decisions

Four indicators, including entertainment (X1.1), interaction (X1.2), trendiness (X1.3) and customization (X1.4), are used to describe social media advertising factors. Customers at Citra Florist Surabaya may be more influenced by social media advertising and the effectively it is managed. Table 12's statistical analysis shows that social media advertising has a positive and significant impact on buying decisions. The quality of a consumer's buying decision increases with the effectiveness of social media promotion. The original sample value (O), which is positive and indicates a direct (positive) influence between variables, is 0.409; the t-statistic value is more than the t-table value (1.645), which is 2.030; and the p-values are lower than 0.05, which is equivalent to 0.043. H1 in this study is therefore acceptable. The social media advertising variable's value in testing the hypothesis is the least in comparison to the other factors. Despite its low value, the social media advertising variable still influences purchase decisions, but less strongly than the other variables. These findings are also in line with [9], which claims that social media advertising enables the development of social interactions that are more intimate and dynamic than those possible with conventional advertising techniques. According to [10], promotion is one of the aspects that determines the success of a marketing program to spread awareness about a product's existence. Based on the value of the loading factor on this variable, impact is the principal indicator (X1.2). Advertising on social media has the potential to affect how consumers decide what to buy. The findings of this study are in line with research [11], that came to the conclusion that social media promotions have a favorable and significant impact on customer purchase decisions and that these decisions are influenced by social media promotion techniques.

The Effect of Word of Mouth on Purchasing Decisions

The word of mouth variable is explained by 3 indicators, namely, the ability of consumers to talk about positive things about quality (X2.1), recommendations for company services and products to others (X2.2), encouragement to friends or relations to make purchases of products and corporate services (X2.3). Based on the results of statistical calculations in table 12, it can be concluded that word of mouth has a positive and significant effect on purchasing decisions. It can be seen from the original sample value (O), which has a positive value, showing a direct (positive) influence between variables, namely 0.457, the t-statistic value which is greater than t-table 1.645, which is 2.327 and the p-value which is smaller than 0.05 which is equal to 0.020. H2 in this study is therefore acceptable. The word of mouth variable appears to have a significant impact on purchasing decisions as evidenced by the highest result obtained when testing the hypothesis for this variable. The degree of factor loading on this variable indicates that encouraging friends or family to purchase the products and services of the company is the most important indicator (X2.3). The results of this analysis support the assertion [12], that word-of-mouth advertising continues to be the most influential marketing tactic in consumers' decisions to buy any given product, regardless of the size of the company. The findings of this study are also consistent with research published in 2019 in SWA magazine and Bee Marketing Research, which shows that the quantity of WOM Dialogue (telling others) makes WOM a source of knowledge for influencing decisions.

Based on the results and discussion that have been described in this study, it can be concluded as that:

• According to statistical analysis, social media advertising has a favorable and considerable impact on purchasing decisions. The value of the original sample (O), which is positive and indicates a direct (positive) influence between variables, is 0.409; the p-values are smaller than 0.05, which is equivalent to 0.043; the t-statistic value is more than t-table 1.645; and the t-statistic value is 2.030. H1 in this study is therefore acceptable. In the Social Media Advertising variable, it might be advised that businesses engage in promotions that are even more alluring in order to persuade customers and increase purchasing decisions. This approach can be implemented by, for instance, including reviews from consumers who have bought the product or by including customer testimonies from each customer

• According to statistical calculations, it can be said that word-of-mouth influences purchasing decisions in a favorable and significant way. The original sample value (O), which is positive and indicates a direct (positive) influence between variables, is 0.457. Moreover, the t-statistic value is greater than the t-table's 1.645, or 2.327 and the p-value is less than 0.05, or 0.020. H2 in this study is therefore acceptable. It might be wise for businesses to keep ongoing promotional initiatives in the word-of-mouth variable in order to boost floral bouquet sales

1. Arifin, Ali. Viral Marketing: Konsep Baru Berinvestasi Dana Berwirausaha. Andi Publisher, 2003.

2. Nasrullah, R. Media Sosial: Perspektif Komunikasi, Budaya, dan Sosioteknologi. Simbiosa Rekatama Media, 2015.

3. Handoko, T. and B.S. Dharmmesta. Manajemen Pemasaran: Analisis Perilaku Konsumen. BPFE, 2012. 

4. Qadhafi, N.E. Pengaruh WOM (Word of Mouth), Harga, dan Label Halal terhadap Keputusan Pembelian pada Produk Air Mineral Dzakya. Skripsi, Institut Agama Islam Surakarta, 2017.

5. Kumala, O.B. Pengaruh Word of Mouth terhadap Minat Beli Konsumen pada Tune Hotels Kota Bali. Fakultas Ilmu Sosial dan Ilmu Politik, Program Studi Administrasi Niaga, Depok, 2012.

6. Sujarweni, V.W. Metodologi Penelitian. 2020.

7. Ghozali, I. Aplikasi Analisis Multivariate dengan Program IBM SPSS. Universitas Diponegoro, 2012.

8. Ghozali, I. Structural Equation Modeling: Metode Alternatif dengan Partial Least Square (PLS). Badan Penerbit Universitas Diponegoro, 2014.

9. Hermawan, A. Komunikasi Pemasaran. Erlangga, 2012.

10. Tjiptono, F. and Diana. Pemasaran. Andi Yogyakarta, 2020.

11. Singgih. Statistik Parametrik. Revised ed., Elex Media Komputindo, 2014.

12. Sumardy. The Power of Word of Mouth Marketing. Gramedia Pustaka Utama, 2011.