Social Media Measurement Widgets And Applications

Social Media Measurement: Widgets and Applications Measureable outcomes of social media engagement (tying revenue to social media actions)

Marc M. Sanford, Ph.D. January 2009

Introduction Social Influence Marketing (SIM), including viral marketing, is occupying significant parts of clients’ advertising campaigns. As such, SIM is also garnering a large amount of advertising dollars. But how much are our clients getting out of this investment? What does successful engagement look like with social media applications? We should not presume that engagement with widgets and other social media applications is unidirectional with revenue.

Problem: Engagement with social media is poorly understood and measured. Little if any research has been done on understanding social media usage and consumer behavior. Solution: Tag and track social media applications, and link users’ behaviors to actions and the consequent conversions on the client site. Benefit: Depending on the purpose of the widget or application, clients can understand the relationship of engagement with social media and the resulting conversion. This adds to the evidence on social media measurement.

Our work here is motivated by our clients. Repeatedly they ask us the following questions: What is the ROI of a social media campaign? How much should I spend on this stuff? Is my piece of social media something people want to share? Is it generating valuable behavior? Do people who get information (widgets, advertising, etc.) from peers act differently than those who are exposed to or find it organically? Because engagement has become a buzzword in our industry, it is necessary to understand exactly what engagement is and how it operates across different platforms. Here we focus engagement within the confines of social media applications, and strive to answer client and industry questions.

Framework & Hypothesis for Assessing Social Media Razorfish has developed a frameworki to measure and assess the success of applications and other components of SIM that have a viral component. An underlying principle to our framework is a hypothesis that those who are more engaged with the application are more likely to convert higher revenue amounts than those who are less engaged. While those who are exposed to social media applications through friends are more likely to convert, we further hypothesize within each of the exposure groups (from friend versus from media or seeding strategy) that those who interact with the social media application are more likely to convert.

© 2008 Razorfish™, Inc. All rights reserved. Razorfish is a registered trademark.

Viral Media Measurement

1

 

Examples In order to test out our assumptions around social media engagement, and to assess the value of viral media, we collected data on two social media applications. Client 1 released a widget that existed on a landing page. Media and friend referrals drove traffic to the landing page where individuals had the option to download the widget or click through to the client site. Client 2 released a widget though Gigya’s distribution network. The initial seeding strategy included roughly 30,000 seeds, and the widget could be shared virally through friends.

Client/Widget Backgrounds Client 1 designed and paid for the seeding of a desktop application. This application resided on a landing page, and traffic was driven to the site through standard media placements. The viral or social component to the application was provided in a link that allowed users to invite friends to download the application. Because this is a desktop application, the seeding component of the analysis for Client 1 treated cookies that arrived via media as paid seeds. The analysis period was from November 27, 2007 through January 15, 2008. Client 2 created and distributed a widget through a social media distribution company that worked with major social networks such as MySpace and Facebook. The campaign ran from February through April of 2008. The widget was used to promote a particular product release, and it included a media player and five areas (tabs) in the widget for interaction. The widget had several functions including the media player, product section, store location area, a “send to friend” option, and a section that allowed the user to post the widget to a social network. It also included a countdown timer to the release of the product. Because of the way the widgets were tracked for Client 2, we were not able to test for differences between those who received the widget from a friend versus those who received the widget virally. Analysis of Paid Seeds For Client 1, 41,599 people arrived through media to the landing page. These individuals placed 744 orders, with an average order value of $175, and they viewed 14,279 unique pages at the site. The total spend for those who arrived via media was $132,257. Additionally, 3,453 people arrived through media and downloaded at least one of the applications. This group generated 56 orders with an average order value of $160, and they viewed 1,310 unique pages on the website. Total revenue generated by those who arrived through media and downloaded the application was $10,307. For Client 2, they paid for 13,000 initial seeds and received 13,388 installs. Approximately 85% of all widget installs were on MySpace. The 13,388 installs generated revenue of $27,209 on the client’s site.

Viral Media Measurement

2

 

Analysis of Widget Engagement Client 1 In the case of both widgets that Client 1 released, each showed significant differences between groups in their level of engagement as tied to unique pages views and dollars spent on the client site. Chart 1: Time spent with widget resulting in average client site unique page views 29

5+ minutes 1‐5 minutes

3

6

8

< 60 seconds

52

Widget2 Widget1

11

Chart 2: Average revenue generated per user by length of engagement with widget 5+ minutes 1‐5 minutes < 60 seconds

$12.96 

$37.33 

$‐ $‐ $1.69  $1.61 

As shown in charts 1-2, additional time spent using the widget resulted in more unique pages the user viewed on the client-site, and a higher overall spend. There is a dip in the number of unique pages viewed, and the dollars spent, in the middle time category. While there may be several explanations for this dip, the general trend is still believed to hold that greater engagement (measured as time spent) results in greater engagement (unique page views) and dollars spent on the client site. Client 2 Table 1: Client 2 Average Revenue per Cookie Engagement is measured through two by Level of Interaction with Widget primary means. The first is the amount of Avg Rev per time that a cookie interacted with the widget. Uniques Revenue Cookie We found that users who were more Viewed 0 Tabs 24% 3% $0.21 engaged with the widget by interacting with Viewed 1 Tab 48% 31% $1.23 more areas (tabs) of the application Viewed 2 Tabs 10% 16% $3.01 accounted for a higher average revenue per Viewed 3 Tabs 8% 9% $2.15 cookie. As shown in Table 1, 48% of Viewed 4 Tabs 7% 29% $8.00 unique users who interacted with only one Viewed 5 Tabs 4% 13% $6.70 tab accounted for 31% of overall revenue, while those who interacted with four tabs accounted for 29% of overall revenue. This results in a higher average revenue per user as interaction increases.

Viral Media Measurement

3

 

The second way we measured engagement was through time spent using the widget. Roughly 72% of all cookies spent less than one minute with the application, while 12% spent between one-to-five minutes, and 16% spent more than five minutes with the application. Despite the large percentage of cookies that spent less than one minute with the application, one cannot characterize these short engagements as poor. The reason being that the average time spent on a website home page is seventeen secondsii. Consequently, the highest revenue value comes from those who spend five or more minutes with the widget. As shown in Chart 4, those who spend five or more minutes with the widget on average spent $9.79 on the client site versus $.35 for those who spent less than one minute. Chart 4: Average revenue per unique cookie by time spent with widget 5+ minutes

$9.79 

1‐5 minutes < 60 seconds

$0.96  $0.35 

Effect of Social We also found significant differences in both engagement and spend between those who discovered the application or widget through media, versus those who were referred by friends. As shown in Chart 5, those who discovered the application via a friend were almost four times more likely to download the application. They were also more likely to spend money on the client site and spent much more on average. More in-depth analysis is reported in the aforementioned report.iii Chart 5: Impact of Social referral on download of widget, spending money, and amount spent 8%

Download Spent $ Amount Spent

23%  Media

1%

9% $3.14 

Friend $23.00 

Total Estimated ROI The estimated ROI for each client varied. For Client 1, the ROI for the social media aspect of their overall campaign was extremely profitable at 377%. The client spent roughly $35,000 on the social media campaign, and earned a return of $132,257 over a two month time period. For Client 2, the results were not as successful with a negative 22% ROI for their campaign. The client spent roughly $37,500 while earning $27,209 of revenue directly from the campaign. There are several reasons why the ROI for this campaign was negative. First, the product promoted had extremely limited quantities (item sold out in about one minute). This means that not everyone who wanted to purchase it had the opportunity to do so. Second, the product was very expensive (retailing for roughly $399). Finally, despite having several unique features (like an MP3 player), the widget was not designed to be a shared very well.

Viral Media Measurement

4

 

Campaign Assessments & Conclusion The campaign by Client 1 was a success because it leveraged available forms of communication with potential customers, and created a large return on investment. However, the viral nature of the widget itself, and the landing environment, were not well constructed as pieces of social media. There were only 216 individuals who arrived at the landing page as a result of a friend sending them an email. Of those 216, only 37 opted to download either of the widgets. One of the fundamental flaws of the widgets themselves was that the user could not choose to share the application after leaving the initial download page. The campaign by Client 2 was successful if you looked at the widget functionality itself. The widget received recognition for its design. However, the widget lacked an aspect that made it truly social, which resulted in low click to share numbers. Both campaigns were successful in engaging the user. The average length of time that users engaged with the social applications gave both clients increased exposure to potential customers. This increased engagement with users resulted in a higher spend on the client site, as well as with a higher engagement with the client site as shown by increased unique page views. From the numbers presented here, one could suggest that social media may be used effectively as a way of engaging users and potential customers. The examples also show that social media may be a cost effective platform for driving users towards conversion behavior. Regardless, the results here are viewed as positive, but with limitations due to small sample sizes. Next steps should include analysis of multiple clients, and establishing baselines for behavior and conversion by vertical. About the Author Marc Sanford is a consultant with the Strategy & Analytics Group for Razorfish, representing a international group out of the Seattle office. Marc works to solve clients’ most pressing issues through evolving strategies and measurement frameworks combined with both quantitative and qualitative analysis. Marc most recently led a team to develop and file a patent for the Generational Tag that tracks viral generations of pass along for social media widgets and applications. With this new technology and through other efforts, Marc is pushing knowledge development in regards to social media measurement and analytics. Marc works with clients such as Best Buy, Nike, Disney, Levi’s, Coors, and many others. Prior to joining Razorfish Marc earned his Doctorate at the University of Chicago in Sociology and taught at the University of Maryland. His dissertation, ”Consumption and the Urban Milieux: Using Consumption as a Measure of Similarity for Defining Urban Neighborhoods” demonstrated how products and consumption patterns within everyday life may be used to define social and geographic groups. Marc joined Razorfish in 2007.

Viral Media Measurement

5

 

About Razorfish™ Razorfish is one of the largest interactive marketing and technology companies in the world, and also one of the largest buyers of digital advertising space. With a demonstrated commitment to innovation, Razorfish counsels its clients on how to leverage digital channels such as the Web, mobile devices, in-store technologies and other emerging media to engage people, build brand loyalty and provide excellent customer service. The company is increasingly advising marketers on Social Influence Marketing™, its approach for employing social media and social influencers to achieve the marketing and business needs of an organization. Its award-winning client teams provide solutions through their strategic counsel, digital advertising and content creation, media buying, analytics, technology and user experience. Razorfish has offices in markets across the United States, and in Australia, China, France, Germany, Japan and the United Kingdom. Clients--many of them served in multiple markets--include Carnival Cruise Lines, MillerCoors, Levi's, McDonald's and Starwood Hotels. Visit http://www.razorfish.com for more information. Razorfish 821 2nd Avenue, Suite 1800 Seattle, WA 98104 Phone: 206.816.8800 Fax: 206.816.8808                                                                                      i Sanford, Marc. “Viral Media Measurement: Modeling Social Applications.” Razorfish. Anticipated 2009  ii Razorfish internal statistics. 

iii Sanford, Marc. “Viral Media Measurement: Modeling Social Applications.” Razorfish. Anticipated 2009 

Viral Media Measurement

6