When Daily Deal Services Meet Twitter: Understanding Twitter As A Daily .

Previous studies on eWOM marketing mainly focus on its effecton consumer judgment and purchasing behavior. Many researchesconfirm that positive eWOM strategy has a positive impact on theconsumer’s buying decisions [10][23]. Some studies look at thepsychological motivations behind eWOM behavior of thecustomers, which are mostly associated with the satisfactiontowards the product they purchased or the level of trust towards abrand [2][25]. A majority of the eWOM marketing studies arebased on the online reviews and ratings [16][19]. [20] and [21]study the impact of eWOM marketing using the data from emailsystems. It is only in recent years that microblogging serviceshave been explored as eWOM marketing platform [12]. [11]examines the expressions of brand attitudes in Twitter postings.To date, little is understood about the nature of eWOM behavior,or the information sharing behavior, on Twitter although manycompanies are using Twitter for marketing and customer services.2.2 Daily Deal ServicesMany daily deal companies are using Twitter for business. Dailydeal sites offer customers daily discounts, called deals, onproducts or services. The daily deal business model is increasingin popularity, and more than 200 daily deal companies exist in theUnited States as of December 2010. Groupon and LivingSocialare the top two leading companies in the U.S. As an example,figure 1 shows screenshots of Groupon. The company provides adetailed description of the deal on the website as shown in figure1(b), and sends the deal information to subscribers through emailas shown in figure 1(a). Daily deal sites are often called groupbuying sites, since a minimum number of purchases have to bereached in order for the discounted offers to become active. Therequired number of purchases is called a tipping point for the deal.To encourage customers to be a part of group buying, users canshare the daily deal information with other users.Prior studies on daily deal services mostly focus on theprofitability of the group buying promotion, and whether it wouldbe effective for affiliated merchants to run promotion throughdaily deal sites in the future [7]. A recent study presents a modelthat can be used to predict the sales outcome of a Groupon dealgiven the deal attributes [13]. By examining the effect ofFacebook ‘Like’ buttons on Groupon sales performance, theauthors provide evidence that daily deal sites benefit from eWOMmarketing on Facebook. Another popular way of sharing dealinformation is by posting URL links of the deals on Twitter,which can be viewed as a type of eWOM activity. Posting URLlinks through Twitter is an explicit action which openly reveals auser’s intention to share information, while Facebook ‘Like’s aremuch more implicit in nature. Furthermore, Twitter differs fromFacebook in that relationships can be uni-directional, allowingone to exchange information with a wider range of other users. Tothe best of our knowledge, there has been no attempt to study theeffectiveness of posting the URL links through Twitter. As shownin figure 1(b) and 1(c), daily deal sites provide buttons for users toshare the URL links through social networking sites like Twitterand Facebook, as well as through email. Investigation of URL linksharing on Twitter can help one understand the eWOM behaviorand characteristics of the daily deal customers, or the customers ingeneral. What kinds of deals are spread, when, and how arecommon questions held by business providers, and understandingof the customers would lead to businesses’ better marketingstrategies.3. RESEARCH PROBLEMSIn this study, we characterize the daily deal information sharingphenomenon on Twitter, with a focus on the spreading of tweetscontaining the URL links that reference the daily deal pages.From now on, we will refer to the sharing of tweets containing theURL links to daily deals as ‘Twitter Sharing’ for the ease ofexpression. We analyze the frequency, range, and timing of thedaily deal sharing to address each of the following questions:1. What are the characteristics of the daily deals shared onTwitter?When choosing to share particular daily deal links on Twitter,customers may be attracted to certain properties associated withthe deal, such as the price of the deal, discount rate of the deal,category of the deal, location of the deal, or whether the deal isoffered in a limited quantity or not. A previous study [13] whichexamines the relationship between the price of the deal and thequantity of the deals purchased by the consumers suggests thatuser demands are price inelastic in case of daily deals, since thedeals already offer heavy discounts. If Twitter Sharing representsthe demand of users, we expect that it would not be dependent onprice nor discount rates. However, we expect Twitter users to besensitive to the limitedness in quantity and locality of the deal; weanticipate that the deals offered in a limited quantity would be of abigger interest to the general public, and the deals available inmultiple locations would be shared by a larger number ofcustomers compared to local deals.2. When are the daily deals shared on Twitter?Understanding of customers also includes awareness of the timeof the information sharing. The volume of tweets containing thedaily deal URL links may vary across different time periods. Aprevious report [4] on the general usage pattern of Twitter claimsthat tweeting activity appears to be predominant during theweekdays, and that it occurs most frequently right after noon. Weexamine whether the deal sharing on Twitter displays similarpattern to the overall Twitter usage pattern, and if daily dealsharing happens more frequently during certain time periods.3. Through what channel are daily deals shared on Twitter?Users can generate tweets through a variety of channels such asthe Twitter website, Tweet Button tools, mobile websites, mobileapplications, and SMS. We investigate through which channelsusers are primarily involved in Twitter Sharing, in order toidentify a target market for businesses to concentrate theirmarketing efforts.4. Is the daily deal sharing on Twitter correlated with the actualsales performance of the deal?We verify whether any positive correlation exists between TwitterSharing and the actual sales performance of the daily dealbusiness. A recent study [13] performed an analysis using thenumber of Facebook ‘Like’s on Groupon deals, and claims thatsocial sharing does have a positive relationship with the daily dealsales. Here we investigate the impact of Twitter Sharing onGroupon sales, and validate whether social sharing on Twitter alsohas a positive relationship with the daily deal sales result.

To address the research problems, we crawled the publiclyavailable data from Twitter, Groupon, and LivingSocial. In thissection, we discuss the details of the collected data and themethods to collect the data.4.1 Twitter DataFor our analysis, we collected a sample of tweets that contain theURL link ‘groupon.com/deals/’. We built a Twitter crawler usingthe streaming Application Programming Interface (API) providedby Twitter 1 , and collected the tweet stream in real-time fromSeptember 3rd 2011 to January 31st 2012. We could only collectpartial data on September 27th due to a technical difficulty2. Thefilter method of Twitter Streaming API returns public statuses thatmatch the designated keyword. Results are given in the JSON(JavaScript Object Notation) format, and we parsed out thenecessary data including the id, text, created date/time, location ofthe tweet as well as the hashtag, mention, and URL contained inthe tweet, if there exists any. A Groupon deal is represented by aunique id following the URL address such as “http://www.groupon.com/deals/ deal id ”. Thus we were able to identify andcollect the tweets which contained links to Groupon daily deals.In total we collected 87,027 tweets written in English.4.2 Groupon DataIn order to study the landscape of the daily deal links sharedthrough Twitter, we collect daily deal information from Groupon,the biggest daily deal site in the United States. Groupon alsoprovides its own API3 for developers to easily access and retrievedeal information. Detailed information about a deal could beobtained by specifying the unique id associated with the deal. Wecollected the information of all the deals that were referenced byat least one of the tweets collected throughout the 5 months period.The results are given in either JSON or XML format, and weparsed out deal id, the original price, discount rate, discountedvalue, category, location, time zone, start and end date of the deal,the total number of the deals sold, as well as whether the deal isavailable in a limited quantity or not. Some daily deal informationwas no longer available after the sales, but the number was fairlynegligible (less than 2% of the total number of collected tweets).After removing the corrupted URL links, we collected 26,804deals referenced by the tweets during the 5 months period.API for developers. Therefore we had to crawl the html pages ofLivingSocial deals in which the necessary data were contained. ALivingSocial deal is represented by a URL address followed by acity id and a unique deal id such as “http://www.livingsocial.com/cities/ city id /deals/ deal id ”. In total we collected 1,069LivingSocial daily deals from 12 different cities in the UnitedStates.5. DAILY DEAL SHARING ON TWITTERIn this section, we discuss the results of the analysis on thecollected tweets containing the daily deal URL links. We explainthe overall trends of daily deal sharing on Twitter, including whatkind of deals are shared when and how.5.1 What Kind of Daily Deals are Shared onTwitter?5.1.1 Daily deal distribution by Twitter SharingFrequencyWe collected 87,027 tweets containing URL links of 26,804different daily deals over a 5 months period. Figure 2 shows thedistribution of the daily deals by the number of tweets thatreferenced the deal. A deal that was shared most often throughTwitter was referenced in 913 different tweets. This is an extremecase since 99.7% (26,736) of the daily deals were shared less than50 times on Twitter. Focusing on the 99.7% of our sample, wedraw a histogram of the set of daily deals that were shared throughTwitter at least once and less than 50 times as well as thecomplementary cumulative distribution curve of the daily deals.frequencycomplementary cumulative distribution curve3000025000# of daily deals4. DATA COLLECTION200001500010000500004.3 LivingSocial DataWe use LivingSocial data to study whether the number of times adeal is shared on Twitter and the actual sales volume of the dealare correlated with one another. We collected the sales volumedata, or the purchase quantity of a certain deal, for LivingSocialdeals. We also recorded the number of times the deals were sharedvia Twitter. In analyzing the correlation we use LivingSocial datainstead of Groupon data for following reason. Groupon stoppedproviding the precise number of purchase quantities as of October2011, because of the external attempts to estimate its total sales.Instead Groupon provides an approximated purchase quantity,such as ‘1000 ’ and ‘5000 ’, causing inaccuracy in our analysis.Unlike Twitter or Groupon, LivingSocial does not provide any1https://dev.twitter.com/2A modification was made to the data transfer protol of TwitterStreaming API on September 29th 2011, and we had to makecorresponding changes to our # of tweets in Twitter Sharing(# of tweets containing the URL link to a daily deal)Figure 2. Daily Deal Distribution by Number ofReferencing TweetsThe bar describes the frequency, and the line is a complementarycumulative distribution curve. We can observe a power lawdistribution (y 64760x-1.796, R² 0.987), with the number ofdaily deals varying as a power of the number of tweets containingthe URL link to a daily deal. Power-law distributions are observedin an extraordinarily diverse range of phenomena [8]. Thedistribution denotes that there are many daily deals that are sharedin only a few tweets, and a few daily deals that are shared in manytweets.We did not plot the daily deals that were shared through Twittermore than 50 times for a better visualization of the power laweffect; the rest not shown on the graph, which account for 0.3% ofthe entire sample also follows the long tail distribution pattern.Only 70 out of the 26,804 daily deals were shared in Twitter morethan 50 times, and 23 deals were shared more than 100 times.

Table 1. Summary of the Top 10 Most Frequently Shared DealsDeal id# ofTweetsGeneralCategorySpecificCategoryAvailable inLimitedQuantityRedemptionLocationDiscountedPrice ( )DiscountRate (%)Descriptiongx-aa-national913ShoppingClothing SalesTRUEMultiple Locations2550American Apparel Discount Coupongx-nat-old-navy-us775ShoppingClothing SalesTRUEMultiple Locations1050Old Navy Discount oney587ShoppingClothing SalesTRUEMultiple Locations1050Old Navy Discount Coupon368ShoppingClothing SalesTRUEMultiple Locations1050Old Navy Discount Coupon349Department StoresTRUEMultiple Locations2050Body Shop Store Discount CouponMovie TheatersTRUEMultiple Locations660Movie Discount Coupongx-gt-donorschooseorg236CharityTRUEMultiple Locations1080Charity Donationga-f-rose-tours229ShoppingArts EMiami82940Backstreet Boys Cruisebeer-and-wine-hobby-1222Food & DrinksHome BrewingTRUEBoston3550Wine & Beer Brewingthe-meat-house-boston218Food & DrinksGrocery StoresFALSEMultiple Locations1050Premium Meat Butcher Shop342We now observe if there are any distinguishable features thatcharacterize the daily deals at the upper tail, or the deals that areshared extremely often on Twitter. Table 1 shows a list of the top10 deals that were most frequently referenced in the tweets.Before discussing the characteristics of the top deals, let usdescribe how the columns are defined. Each daily deal belongs toa general category and a specific category predefined by Groupon.Some of the deals are available in a limited quantity and some arenot. Most of the deals on daily deal sites are local deals that areonly available at one specific location, while some of the deals,including the online deals, are available at multiple locationsaround the nation. For all deals, the original price and thediscounted price along with the discount rate information areprovided. Brief descriptions of the deals are presented in the table.Several common characteristics could be observed among the top10 deals. First, all but one deal are limited in quantity. The goal ofselling products or services in a limited quantity is to boost upcustomers’ interests in the product and motivate them to consumea product by creating a sense of urgency. Previous researchesclaim that such marketing tactic, also known as scarcity marketingor exclusivity marketing [17][18], is effective. This strategy isused by many companies, one of them being Groupon. Anobservation of the top deals suggests the need to examine whetherthe degree of Twitter Sharing of the deal differs upon the limitedavailability of the deal, which will be discussed in Section 5.1.4.Second, the top 7 deals offer discounts that can be redeemed atmultiple locations. For example, the most frequently referenceddeal corresponds to a voucher that could be used in AmericanApparel, a clothing brand in the United States. The voucher wasvalid at all U.S. stores or at the online store, attracting a largenumber of users to make purchases and post tweets. This leads toa need to examine the difference in the degree of Twitter Sharingdepending on the availability of the deals in multiple locations.Furthermore, a breakdown of the daily deals shared on Twitter bylocation would help businesses to understand the userinvolvement in social sharing across different cities and time zone.We conduct analysis to address these questions and present theresults in Section 5.1.3.Third, the top 5 deals belong to the shopping category. Users mayexhibit dissimilar Twitter sharing behavior for deals fromdifferent categories due to varying demands across differentcategories. A breakdown of the daily deals shared on Twitter bycategory would help one understand the different extent to whichpeople are interested in sharing the information. We conductfrequency analysis of the tweets containing the daily deals acrossdifferent categories in Section 5.1.2.Now we take a look at the price and discount rate of the top deals.The prices of the top deals vary from 6 to 829 dollars, and nogeneralization could be made in terms of price. In economics, it isargued that the number of consumers interested in acquiring theproduct rises as the price falls by the law of demand. In otherwords, the demand of customers are said to be price elastic.However, a recent study [13] argues that this does not apply to thecase of daily deals; the demands of daily deal customers areshown to be price inelastic because daily deals already offer largediscounts. We expect that the price or the discount rate of a dealwould not be related to the degree to which it is shared on Twitter,under the assumption that Twitter Sharing closely reflects the userdemand. We find it necessary to conduct a quantitative analysis toverify whether the assumption is true, and discuss the analysisresults in Section 5.1.5.Other interesting facts from the data worth noting are as follows:The 7th most frequently shared daily deal on the list was a charityevent, encouraging people to donate funds to an online charityorganization named DonorsChoose.org. At times, daily deal sitesteam up with non-profit organizations to promote nationwidecharity drives and to raise public awareness. Our data reveals thatthe links to such charity events are frequently shared on Twitter,indicating that the explanation for information sharing is partiallyrooted in altruistic purpose. Another interesting fact could beobserved from the 9th deal on the list. Interestingly, many Twitterusers shared URL links to this deal despite the fact that it was alocal deal. An additional examination of the tweet contentsrevealed that this was largely due to a giveaway event hosted bythe local marketer. A previous study [15] argues that one of thereasons why Twitter users remain subscribed to businesses andcorporates is because they enjoy the giveaways and other eventsoffered on Twitter by the businesses. Our data also shows that thegiveaways and other events have the power to increase publicparticipation on Twitter.5.1.2 Daily Deal Distribution by CategoryIn this section, we present a frequency distribution of the Groupondeals shared on Twitter across different categories. All the dealsfall into one of the 18 general categories predefined by Groupon:‘Arts & Entertainment’, ‘Automotive’, ‘Beauty & Spas’,‘Education’, ‘Financial Services’, ‘Food & Drink’, ‘Health &Fitness’, ‘Home Services’, ‘Legal Services’, ‘Nightlife’, ‘Pets’,‘Professional Services’, ‘Public Services & Government’, ‘RealEstate’, ‘Religious Organization’, ‘Restaurants’, ‘Shopping’, and‘Travel’. By collecting all the tweets that contain the links to thedaily deals and the corresponding daily deal information, we wereable to observe the different degree to which the deals of differentcategories were being shared on Twitter.

Figure 3 illustrates the distribution of the deals across differentcategories. The plot shows the frequency of Twitter Sharing forthe 12 most freque

2.2 Daily Deal Services Many daily deal companies are using Twitter for business. Daily deal sites offer customers daily discounts, called deals, on products or services. The daily deal business model is increasing in popularity, and more than 200 daily deal companies exist in the United States as of December 2010. Groupon and LivingSocial