COINs 2013 Course Project 9 Report
Exploration of Methods to Measure Emotional Responses to User Interfaces: A Case Study on Citrix’s GoToMeeting Emanuel Castillo, Priscilla Mendoza1*, Bhavika Shah2*, Yulia Tammisto University of Cologne, Albertus-‐Magnus-‐Platz, Cologne 50923, Germany Savannah College of Art and Design, 3116 Montgomery Street, Savannah 31401, USA Aalto University School of Business, Runebrtginkatu 14-‐16, Helsinki 00100, Finland
Abstract Online social communities are an invaluable avenue for understanding user emotions. In this paper, our collaborative innovation network (CoIN) describes our attempt to find methods suitable for measuring user emotions through sentiment analysis of online data. Our aim was to develop a methodology that could help companies use customer reviews and posts from social media platforms to spot insights for improving product strategy and thus performance. In order to conduct the analysis of the data, we tested a number of social network data gathering and sentiment analysis tools to be able to choose the few that were the most suitable for measuring emotions. During testing, we quickly saw the shortcomings of the quantitative sentiment analysis tools and therefore searched for sentiment analysis tools that would analyze qualitative data in order to capture the full meaning of each review and post. The findings developed into new opportunities specifically for Citrix’s GoToMeeting product team, but can potentially be expanded to other companies using social network analysis and sentiment analysis on their customers’ online reviews and posts to understand and process their information. Keywords: Social Network Analysis, Sentiment Analysis, Qualitative Sentiment Analysis, User Interface, Emotions, Corporate Identity, AC2ID Test
* Corresponding author. Tel.: +1-‐912-‐306-‐6919 E-‐mail address: firstname.lastname@example.org * Corresponding author. Tel.: +1-‐281-‐217-‐0019 E-‐mail address: email@example.com
1. Introduction Our research began as an attempt to help companies better understand and connect to their customers through social media. In order to do this, our study focused on the video conferencing software GoToMeeting by Citrix. Citrix is a multinational software company that develops desktop visualization, networking, software as a service and cloud computing solutions. GoToMeeting is video conferencing software that helps customers participate in online meetings and includes functions as desktop sharing, one-‐click recording and HD faces. GotoMeeting transmissions are protected with high-‐security encryption and optional passwords. Thus, sales are mostly focused on b2b markets such as employees of companies that have purchased GoToMeeting. Their motto is “online meetings made easy.” Citrix’s management was particularly interested in the emotional reactions of their end users to their software interfaces. This would be possible through Social Network Analysis and possibly discovering the potential to develop improvements for their product. Therefore from the very beginning, our team’s goal was to provide practical results that could be applied to improve business performance for the company. We began by developing a project plan that would help us to deliver implementable yet suitable results for application by Citrix. The initial goal of the project was as follows: To research users’ emotional reactions to an interface design through social media. However, after conducting initial research, we saw that there was an opportunity to understand and analyze users’ emotional reactions to spot innovation opportunities for products and companies. Our plan began by collecting discussions, comments, and reviews about GoToMeeting by end users and experts on various social media platforms to perform a sentiment analysis. We also gathered blog posts, updates, and reviews posted by Citrix and GoToMeeting, and ran the same sentiment analysis. By doing both, we were able to compare the company’s emotional appeal to the product with users’ emotional response to its interface. The comparison helped us understand and describe the gap between the company’s self image and the way customers perceive the company. Knowing and understanding this gap, in turn, would help Citrix address the product development and communications accordingly and to potentially minimize the disappointment and negative effect caused by the mismatch.
2. Data Collection Channels According to our plan we started gathered two types of data: “internal” from Citrix about GoToMeeting and “external” from GoToMeeting’s users and tech experts reviewing and posting their experience with the software. The sources were searched throughout the Internet. We used the software Condor to help us identify the websites that mention and post about GoToMeeting. Combined with Google Search, this gave us a wide range of different sources including: Citrix’s corporate pages, their blog, and their Facebook profile; various tech expert blogs; and websites collecting user reviews on the product (including Google Play, Amazon.com, AppStore and other smaller resources.) We were also able to obtain data from Citrix’s management about their vision of the product that we used for analysis. In order to be able to analyze the data in our project timeframe, we limited the data gathered to between 2011 to 2013. A short description of each source group is presented in the following sections.
2.1. Citrix Corporate Pages We collected the mission and vision statements and product descriptions for GoToMeeting from Citrix’s corporate website (citrix.com), and data for analysis of Citrix’s image were gained through reviews from Citrix website (citrixonline.com). In the corporate blog (blogs.citrix.com) we also gathered updates and news about GoToMeeting. Also, we collected all GoToMeeting related reviews and manually copied the content into an Excel file. We copied other words found on the website such as the text in sidebars in order to understand the image and message that Citrix Online wants to communicate.
2.2. Facebook – GoToMeeting Page Facebook has about 1.19 billion monthly active users (https://newsroom.fb.com/Key-‐Facts (01-‐23-‐2014)) and therefore plays a big role in the area of social media analysis. For the purpose of our work, we collected user posts on the GoToMeeting Facebook Page (https://www.facebook.com/GoToMeeting) from January 2012 to November 2013. This source was selected in order to obtain the unprompted experiences and emotions of users that consider GoToMeeting as a “friend.” Posts that weren’t related to the GoToMeeting product, such as posts about Halloween or Christmas, were not collected.
2.3. Blogs and other general websites Our Condor and Google searches revealed a large number of tech-‐industry related pages such as expert blogs or online journal websites that wrote posts and reviews about GoToMeeting. We copied all the material, articles, and reviews to our Excel data file for further analysis in order to learn what those proficient in computer software were saying about their experience with GoToMeeting. The more detailed list of sources is presented in Table 1. Table 1. Blogs and other general websites sources discussing GoToMeeting Source type
Tech Rating Blog
Table 1. Blogs and other general websites sources discussing GoToMeeting (continues) Source type
2.4. Twitter Twitter was one of the most recognizable social media platforms to find emotional user responses and measure sentiment. We collected and analyzed Twitter data concerning GoToMeeting and Citrix through tools that gathered and grouped the tweets instantaneously (described in the following section “Methods”). This was the only data we did not have to collect manually. The tweets helped us discover ephemeral yet emotional user responses concerning GoToMeeting and Citrix.
2.5. Markets (Google Play, AppStore, Amazon.com) The digital products markets were a source of multiple reviews about GoToMeeting. The largest database was available through the AppStore (over 1000 entries), since the GoToMeeting app has been sold through that avenue for a few years. However, Amazon has only sold the app for about a year and less than a year on Google Play. We collected the reviews left in 2011, 2012 and 2013 to be able to analyze them in conjunction with development and upgrades of the app.
2.6. Youtube video comments During our search for data about GoToMeeting, we found a number of video comments posted on the YouTube commercials. We decided to include data from this channel since the people left both comments about their experiences with GoToMeeting software as well as critiques about the video. Overall we were able to gather over 100 video comments. Those comments that were not relevant to GoToMeeting or contained random comments were excluded.
3. Methodology To proceed to analysis, we first explored the wide range of different tools and tested them on the data collected. The purpose was to understand how each method worked and to be able to choose the most suitable tool that would help us reach and validate our project goal. Each tool we found and applied to our data is presented in the following subsections including a short description and examples from the data.
3.1. Social Network Analysis tools 3.1.1. Condor Condor is a data collection software developed by Peter Gloor that located the websites that reference Citrix and GoToMeeting. We also ran searches on Wikipedia and Twitter feeds that gave us the most prominent actors that discuss Citrix (Mark Templeton) and GoToMeeting (Keith Ferrazzi). This information was used to locate further resources in our search for data, but not used in analysis.
Figure 1. Print screen of Condor Wikipedia search with GotoMeeting, retrieved 11/04/13
3.1.2. Topsy3 Topsy is a data collecting and analysis software for socially shared content primarily from Twitter and Google Plus. It provides analytics based on the key word counts, key actors and top influencers, sentiments over time, and sentiments according to each tweet. We use Topsy to search and analyze tweets discussing Citrix and GoToMeeting. Some examples of each are presented below.
Figure 2. Print screen of Topsy search for CITRIX and GotoMeeting, retrieved 11/03/13
Figure 3. Print screen of Topsy sentiment score for CITRIX and GotoMeeting, retrieved 11/03/13
Figure 4. Print screen of Topsy top influencers talking about CITRIX and GotoMeeting 6
3.2. Sentiment Analysis tools 3.2.1. IBM Many Eyes4 IBM Many Eyes provides a wide functionality for visualization of all sorts of data, from text to mathematical data. However, we only used the text analysis tools: word clouds, word trees, tag clouds, and phrase net. The word cloud provides a picture of the most used words in our dataset where size and boldness of the font corresponds to frequency (see Figure 5). The cloud tag is an expanded version of word cloud; it calculates the frequencies of each work in a text and visualizes those words as tags. Although on the top of the tag cloud visualization the exact number of each word occurrences and the context in which it was used is shown (see Figure 6). The phrase net diagrams the relationships between different words used in a text. Thus it helps pinpoint and evaluate the meaning behind each word used, in what context the word was used, and what other words linked to it most frequently (see Figure 7). Similarly, the word tree shows a visualization that connects words that were paired together. This way we don’t only evaluate if the sentiment is positive or negative but can track the subject of the message and see what particular features of the service were praised or criticized by authors of the texts we analyzed. Therefore, we were able to check the individual words of interest and their relations within our data, (see Figure 8, the word “need”). By looking at the words “good” and “bad” (and other nouns with similar meaning) we were able to understand the product features that customers were satisfied with and which ones left them feeling dissatisfied.
Figure 5. Print screen of IMB Many Eyes word cloud analysis of Blog posts about GotoMeeting, 2011
Figure 6. Print screen of IMB Many Eyes tag cloud analysis of Blog posts about GotoMeeting, 2011
Figure 7. Print screen of IMB Many Eyes phrase net analysis of iTunes GotoMeeting reviews, 2013
Figure 8. Print screen of IMB Many Eyes word tree analysis of iTunes GotoMeeting reviews, 2013
3.2.2. Lexalytics5 Lexalytics is a software package that uses its own algorithm to calculate sentiment of a given text. It does so through transforming unstructured text into structured data. The software extracts entities (people, places, companies, products, etc.), sentiment, quotes, opinions, and themes (generally noun phrases) from text. The software uses natural language processing technology to extract the above-‐ mentioned items from social media, forums, or general text. The example of using Lexalytics on our data is shown in Figure 9.
Figure 9. Print screen of Lexalytics analysis of iTunes GotoMeeting reviews, 2013 5
3.2.3. LIWC6 Linguistic Inquiry and Word Count (LIWC) is a text analysis software program. LIWC calculates the degree to which people use different categories of words across a wide array of texts. As an outcome, it estimates the degree any text uses positive or negative emotions, self-‐references, causal words, and 70 other language dimensions. The LIWC program analyzes various standardized ASCII text files or Microsoft Word documents. The LIWC2007 licensed program also allows you to build your own dictionaries to analyze dimensions of language specifically relevant to your interests. The student version of LIWC, LIWClite7, we used only analyzes plain text files using the LIWC2007 and earlier LIWC2001 dictionaries. The demo version is also limited to the amount of data you can input at once, so it was rather inconvenient to use for our purposes. The screenshot of the analysis we made with this tool is shown in Figure 10.
Figure 10. Print screen of LIWC analysis of iTunes GotoMeeting reviews, 2013
3.2.4. Wordle7 Wordle is a tool very similar to IBM Many Eyes word cloud function. It has a simple and user-‐friendly interface for quick generation of word clouds that can be customized by color and/or font. Since it did not provide any additional functionality over IBM Many Eyes, we used the latter. However, Wordle is a good tool when word cloud analysis is sufficient and may be preferable because of its simplicity for use and ease of access.
3.2.5. Text-‐processing.com API8 The text-‐processing.com API offers various functionalities for text mining and natural language processing through a JSON over HTTP web service. Two functionalities were used for our purpose, sentiment analysis and word stemming. The website of this tool offers client libraries for different programming languages such as Java, PHP and Python. Since we didn’t need all the various 6
functionalities, we decided to write a short program code in PHP and to import the data and results in a MySQL database. The sentiment analysis of the text-‐processing.com API is composed by two bayes classifier and was trained by movie reviews. The first classifier calculates the probability for a text to be neutral. The second classifier calculates the probability of the text to be negative or positive. Given that the probability of the neutrality is higher 0.5 the final classification would be neutral. If the text is not neutral, this means probability is lower than 0.5, the text will be classified negative or positive depending on which value is higher. Table2. The code running analysis on reviews and other GoToMeeting data in text-‐processing.com API
Figure 11. Print screen of Text-‐processing.com API analysis of a GoToMeeting review
3.2.6. Sentiment.Viz9 Sentiment.Viz is a social media data analysis tool for Twitter. It searches tweets by an inputted key word, in our case – GoToMeeting, and uses a sentiment dictionary (Figure 12) to estimate the sentiment of each tweet. It analyzes the words in each tweet, and then combines the words' pleasure and arousal ratings to estimate sentiment for the entire tweet. The full functionalities of this tool that we used for our analysis are described below: •
Sentiment. Each tweet is represented as a circle positioned by its individual sentiment. Unpleasant tweets are drawn as blue circles on the left, and pleasant tweets as green circles on the right. Sedate tweets are drawn as darker circles on the bottom, and active tweets as brighter circles on the top (see Figure 13). Topics. Tweets about a common topic are grouped into topic clusters. Keywords above a cluster indicate its topic. Tweets that do not belong to a topic are visualized as singletons on the right (see Figure 15). Heat Map. Pleasure and arousal are used to divide sentiment into an 8×8 grid. The number of tweets that lie within each grid cell are counted and used to color the cell: red for more tweets than average, and blue for fewer tweets than average. White cells contain no tweets. Tag Cloud. Common words from the emotional regions upset, happy, relaxed, and unhappy are shown. Words that are more frequent are larger. Timeline. Tweets are drawn in a bar chart to show the number of tweets posted at different times. Pleasant tweets are shown in green on the top of the chart, and unpleasant tweets are shown in blue on the bottom. Map. Tweets are drawn on a map of the world at the location where they were posted. Please note that most Twitter users do not provide their location, so only a few tweets will be shown on the map.
Affinity. Frequent tweets, people, hashtags, and URLs are drawn in a graph to show important actors in the tweet set, and any relationship or affinity they have to one another. Tweets. Tweets are listed to show their date, author, pleasure, arousal, and text.
Figure 12. Sentiment.Viz scale for evaluating different emotions through sentiment analysis
Figure 13. Sentiment.Viz analysis of Twits about GoToMeeting, retrieved 11/03/13
Figure 14. Sentiment.Viz analysis of GoToMeeting with twits view, retrieved 11/03/13
Figure 15. Sentiment.Viz analysis of Twits about GoToMeeting the clustering of results
3.3. Evaluation of sentiment analysis tools After running our data through a range of various algorithms for sentiment analysis, we needed to choose the tools that proved to be valid and helped deliver us to our project goals. In order to narrow the tools down, we created a set of conditions: • •
The first condition: the analysis of all our data used the same or at least similar algorithms. The second condition: determine the quality of the sentiment analysis. The technical ability of each method to distinguish sentiment correctly is limited compared to human analysis, so we wanted to choose the tool that would perform the best quality that could as closely match human analysis. 14
The third condition: expand beyond the classification of positive/negative/neutral sentiment and highlight different emotions that users are expressing. To deliver our project goal – “analysis of user emotional responses to identify the opportunities for further development of the target product/company” we needed to capture a broader range of emotions than “good vs. bad.”
3.3.1. Data input methods of the tools Many of the tools we used did not allow large text or data input. Despite the great functionalities of Topsy and Sentiment.VIZ, we discovered that these tools were only suitable for analysis of tweets. We did contact the developers of Sentiment.VIZ to find out if it could support data from other sources or text input, but this functionality was not available. However, we still could take into account the results derived from Topsy and Sentiment.VIZ but only as supplementary information. The tools with sufficient input options were Lexalytics, word-‐processing.com, IMB Many Eyes, and Wordle. LIWC would have been an option, but there were limitations in how much data could be inputted using the demo version so it was not feasible for our analysis.
3.3.2. Validation of Sentiment Analysis We chose our final evaluation tools according to the availability of inputting large texts and how similar they were to human sentiment analysis. According to our first requirement, the tools were narrowed down to Word-‐Processing.com, Lexalytics, and Qdap (R-‐package). Then, by evaluating the quality of the sentiment analysis and their algorithms, we performed a manual sentiment of a random sample from our dataset (approximately 10% or 100 entries) and compared the results of each tool. We then calculated the Euclidian distance (see Figure 14) to the manual answer from each of the tools values to verify their sentiment analysis and evaluate which tool we should move forward with.
Manual Analysis 100 cases
For validation of the various tools we used, we performed a manual sentiment evaluation. We read 10% (100 entries) of our AppStore reviews and gave each a value of -‐1 if the sentiment sounded negative; the value of +1 if the sentiment sounded positive; and 0 for neutral reviews.
QDap – R package
Quantitative Discourse Analysis Package (QDap) is a statistical software tool originally designed for analysis of transcripts but is also suitable for other purposes such as sentiment analysis. Its functional possibilities are frequency counts of sentence types, words, sentences, turns of talk, syllables and other assorted analysis tasks. QDap was only used as a comparison tool for the manual analysis. QDap is one of the tested analysis tools in Table 3.
Figure 16. Euclidian Distance Calculation We calculated the distances of each individual entry assessed by each tool, gained the final score, and compared each entry to see which one performed closest to the manual sentiment analysis. The comparison is illustrated in Table 3. According to the total distance score, the closest algorithm to manual sentiment analysis was Lexalytics. However, the Euclidian distance validation technique has flaws so we cannot take is an absolute. For example if a tool provides a lot of neutral judgments then the distance would be smaller even if most of the values do not coincide (since the distance from 0 to +1 as well as -‐1 will be the shortest). In order to combat this flaw, we took the cumulative frequencies of each judgment (-‐1, 0, +1) made by each method. The result is summarized in a histogram in Figure 15. Table 3. Comparison of the results of the tools to manual sentiment analysis
Figure 17. Comparison of Cumulative Frequencies of Sentiment Analysis made by different algorithms According to Figure 17, the algorithm of Text-‐processing.com API tool performed the closest to manual analysis. The results of two validation techniques we applied are controversial as they point to different tools thus there is not an absolute winner. Therefore, we decided to use both Lexalytics and Word-‐ Processing.com for further analysis. Finally, using the third evaluation method we pointed out in the beginning of this subsection, the availability of wider scale of emotions, we needed to supplement Lexalytics and Word-‐Processing.com with another tool proceed with further analysis. The most suitable tool that describes emotions on a relative scale is IBM Many Eyes (word tree, word cloud, tag cloud). By using all three together, we were able to build a stronger foundation for our data analysis.
3.4. Theoretical Foundation to Frame the Findings Our plan also consisted of comparing the two sets of data we have collected: “internal” what Citrix is saying about the target product and “external” what customers, users, and various experts are saying about this product. To perform this comparison, we tested and evaluated the data using the three evaluated sentiment analysis tools. However, in order to make solid conclusions based on our findings, we needed a theoretical framework. This framework would help us better understand the interconnections between our constructs and how to interpret our findings in order to make them comprehensible and applicable for management. The Corporate Identity Management theory was used to develop opportunities and insights from our data analysis. Corporate Identity Management theory argues that each corporation has a “character” and best serves customers who share or like similar identity traits (Bromley 2000). Thus it is crucial for a company to be consistent in transmitting its “character” to the market to find the “right” customers. In our research, comparing Citrix’s understanding of its own identity (through the lenses of the target product) to that of the customer’s perception, we could see if there were any identity gaps. The framework built from Corporate Management Theory is defined by the acronym ACID (Balmer & Soenen 1999), where “A” refers to actual identity of a company, “C” refers to communicated identity, “I” refers to ideal identity, and “D” refers to desired identity. The idea of the test is to pinpoint the different identities a company has and compare them to each other. A modification to the ACID Test is the AC2ID test (Balmer & Greyser 2002) that has an additional “C” for conceived identity. After distinguishing the five identities of the AC2ID Test for GoToMeeting, we were able to compare the “internal” and “external” images and formulate the insights and opportunities. 17
Figure 18. AC2ID test model by J.M.T. Balmer and S.A. Greyser
4. Insights As mentioned earlier, after screening our data, we were able to separate our data to into the constructs of the AC2ID Test for analysis. The five subsections below describe each of the identities and attribute them to our analyzed datasets in detail.
4.1. Actual Identity Actual is an identity that defines who the company is, on a structural and strategic decisions level. Balmer and Greyser (2002: 74) state that “the actual identity constitutes the current attributes of the corporation. It is shaped by a number of elements, including corporate ownership, the leadership style of management, organizational structure, business activities and markets covered, the range and quality of products and services offered, and overall business performance. Also encompassed is the set of values held by management and employees.” For this identity, we analyzed Citrix’s mission and vision statements to reflect the actual identity of the company. The resulting analysis of this data is: GoToMeeting corporate content suggests that this application has the objective to be the facilitator for enterprises to collaborate more internally, and for individuals to accommodate work and life in a more flexible way.
4.2. Communicated Identity Communicated identity refers to the image the company is transmitting to outside world, what it tells about itself. Balmer and Greyser (2002: 74) define it as follows: “The communicated identity is most clearly revealed through ‘controllable’ corporate communication. This typically encompasses advertising, sponsorship, and public relations. In addition, it derives from ‘non-‐controllable’ communication, e.g., word-‐of-‐mouth, media commentary, and the like.” For this identity, we analyzed marketing materials found through corporate blog and reviews posted by Citrix. The following is our analysis of communicated identity:
GoToMeeting’s communication emphasizes the collaborative element of working together where in different locations. It also highlights the cost effective results of using this application. The company has made a great effort on collecting testimonials from a variety of clients within different industries.
4.3. Conceived Identity Conceived identity is how users and non-‐users view and evaluate the company, Balmer and Greyser (2002: 74) explained it as follows: “The conceived identity refers to perceptual concepts —corporate image, corporate reputation, and corporate branding. These are the perceptions of the company—its multi-‐attribute and overall corporate image and corporate reputation—held by relevant stakeholders.” In our study conceived identity refers to all the social media data we collected. This dataset is compiled of over a thousand customer reviews and posts, expert opinions from blogs and web, ratings and reviews from app market places (Amazon, Google Play and AppStore), social networks posts (Facebook), popular tech blogs, review aggregation pages, and video comments (Youtube). The outcome of the analysis is summarized below: The user impressions about GoToMeeting were mixed, positive and negative comments were almost balanced in our dataset. The strengths and weaknesses of GoToMeeting app, which customers and other reviewers pointed out the most, are listed below. According to our data the app strengths that evoked positive reactions and comments were • • • •
Flexibility on joining meetings remotely Lifestyle change High Definition of the meeting participants’ faces Availability on mobile devices
The weaknesses of the app that people mentioned the most were related to • • •
Lack of a feature to inform the presenter when a participant has a question (that would refer to raising hand activity in a classroom situation) Lack of possibility to host files A perceived complexity of hosting meetings
4.4. Ideal Identity This identity identifies the best a company can make out of its capabilities. In other words, it is the system perspective of a company’s best position or role in a global ecosystem. The description from the authors of the concept (Balmer & Greyser 2002> 74) is quoted below: “The ideal identity is the optimum positioning of the organization in its market (or markets) in a given time frame. This is normally based on current knowledge from the strategic planners and others about the organization’s capabilities and prospects in the context of the general business and competitive environment. The specifics of a given entity’s ideal identity are subject to fluctuation based on external factors—e.g., the nuclear power industry after Chernobyl; and industries (such as travel, transport equipment, and security systems) affected negatively and positively by the September 11 World Trade Center catastrophe.” 19
In order to evaluate the ideal identity, it would require collecting data from other resources such as news and financial analysts’ reports (that we did not have access) to make reliable judgments. Due to the limitations, we decided to omit this identity from our analysis and proceed with the other four identities that are described. In the following section we make comparisons between those four identities and discuss possibly implications for the Citrix and GoToMeeting.
4.5. Desired Identity This identity is formed by what the management wants the company to be and to what direction they wish to develop their organization. Balmer and Greyser (2002: 75) describe it as follows: “The desired identity lives in the hearts and minds of corporate leaders; it is their vision for the organization. Although this identity type is often misguidedly assumed to be virtually identical to the ideal identity, they typically come from different sources. Whereas the ideal identity normally emerges after a period of research and analysis, the desired identity may have more to do with a vision informed by a CEO’s personality and ego than with a rational assessment of the organization’s.” Our analysis of Citrix and GoToMeeting was developed through our contact of management at Citrix and by the mission and vision statements of Citrix and GoToMeeting. Through discussions and presentations, we formed the desired identity of Citrix: GoToMeeting wants to be THE online meeting software that provides the tools necessary to allow users to feel as comfortable and share information is if they were meeting face-‐to-‐face.
5. Opportunities By developing the various identities of a company, we can determine whether the identities are aligned. In this section, we compare Citrix’s identities described in the previous section and compare how well they correlate. We create pairwise comparisons of selected identities that can potentially help Citrix and GoToMeeting better understand their customers and perform adjustments to its communication, customer service and product development strategies.
5.1. Actual vs. Communicated By comparing our analysis of actual and communicated identities, we found that although GoToMeeting communicates what it stands for in its vision and mission there were still areas for improvement in addressing more niche customers. They can add the focus on B2C activities to its B2B strategy to help improve product development and deliver offerings to individual users along with those for organizations. For example, the user sector of independent agents such as nutritionists or personal and financial coaches have not been yet addressed by the company’s communication. There is an opportunity for establishing contact to this entrepreneurial sector via marketing channels and product development strategy add-‐ons (in features or capabilities of the app).
5.2. Desired vs. Conceived Although, the management sees Citrix as an end user-‐centered company, our analysis discovered that there are still issues to be resolved in order to achieve this state. GoToMeeting thrives for ease of use and practicality, however there are some users that struggle with existing features or require more 20
capabilities to get their job done. Thus there are possibilities in product development that can help Citrix fulfill the vision of management and become a company they are sought to be.
5.3. Communicated vs. Conceived The comparison between communicated and conceived identities revealed that GoToMeeting’s marketing materials generally addressed organizational customers (companies) and the front-‐end users that set up the meetings, whereas the social network data was usually reviewed and rated by end users: employees of the organizational customers and individual customers. Thus we could detect the opportunity for Citrix and GoToMeeting to publish experiences from individuals that are using the software to relate more to end users and help develop the image of a user-‐centric company that Citrix represents.
6. Conclusion and Next Steps GoToMeeting by Citrix is a popular yet young software with thousands of downloads and millions of users. Although a multitude of information could be collected through the web, our method applied to the selected amount of data helped determine new opportunities for further development for Citrix to better understand and connect with their customers. Our research, analysis and insights aimed at understanding user emotional responses to the product through social network analysis. By collecting data from a multitude of sources and analyzing them through various tools, we were able to build a reference for comparing sentiment analysis. By checking the results to the manual sentiment analysis, we validated our findings and expanded beyond the positive/negative score to encompass a broader range of emotions. Finally, the application of the AC2ID Test theoretical framework to our sentiment analysis outcomes allowed us discover insights from the data and create recommendations in which Citrix might embrace to improve user experiences. We hope that our final recommendations validate the use of Social Network Analysis as a strategy for Citrix to build upon for GoToMeeting. As we progressed on our project, we discovered a few opportunities that were not pursued due to lack of time, but that are interesting and could be developed further. For example, there is potential to develop some of the tools we tested, Sentiment.VIZ and Topsy, to expand their functionality to perform qualitative sentiment analysis on all types of texts. We found these tools very useful and rich in their descriptions of emotions compared to the basic three-‐option sentiment analysis: positive, negative, and neutral. In our opinion, there is a great opportunity to develop such software for companies to use to help in a deeper understanding of users through social network analysis. Finally, we believe that social network analysis has the potential to be used as a form of primary research in the design of new or redesigned projects. Users are already giving unprompted information about how they feel about a product or service, so proper analysis of this information can give a company an edge on addressing the needs of their customers.
References Balmer, G.M.T & Greyser, S.A. (2002). Managing the Multiple Identities of the Corporation, California Management Review, 4 (3) : 72-‐86 21
Bromley, D. B. (2000). Psychological Aspects of Corporate Identity, Image and Reputation, Corporate Reputation Review 3 (3): 240-‐252 Balmer, G.M.T & Soenen, G.B. (1999). The Acid Test of Corporate Identity Management™, Journal of Marketing Management, 15(1-‐3): 69-‐92 Freeman, L. (2004). The development of social network analysis: A study in the sociology of science. (1st ed.). North Charleston, South Carolina: Empirical Press. Gloor, P., & Cooper, S. (2007). Coolhunting: Chasing down the next big thing. (1st ed.). New York, NY: AMACOM. Gloor, P. (2010). Coolfarming: Turn your great idea into the next big thing. (1st ed.). New York, NY: AMACOM. Rainie, L., & Wellman, B. (2012). Networked: The new social operating system. (1st ed.). Cambridge, MA: The MIT Press.
Published on Feb 27, 2015
Published on Feb 27, 2015
Online social communities are an invaluable avenue for understanding user emotions. In this paper, our collaborative innovation network (CoI...