What Social Media Ad Types Are Working Best?

Executive Summary.
 Increases in social networking and interactive advertising technology have led to a variety of new options for advertisers and publishers. With the expansion of this medium, choosing the right ad type for a campaign may not always be straightforward. The purpose of this study was to experimentally test user engagement across a variety of so-called “widgets,” defined to be interactive ad types that may prompt users to engage their peers in a social manner. Survey respondents were randomly assigned to view a narrated video of one of 7 ad types, and then report on their engagement, perceptions, and purchase intent. A summary of the findings are as follows: Sponsored Content ads, in which individuals viewed a holiday page that was “brought to you by” a leading brand, were the most engaging but produced the least purchase intent of the 7 ad types tested. Corporate Profiles on social-networking sites produced greater purchase intent and more recommendations when users could become a fan and add the logo to their own profiles than when they could not. Give and Get Widgets in which individuals could create and customize something (a car or a dinner menu) and then either send it to a friend (“give” widget) or keep it for themselves (“get” widget) were more engaging than traditional banner advertisements but no more likely to produce an intent to purchase. The above conclusions held across brands (a leading soup brand and a leading car brand) and publishers (on Allrecipes.com and on Facebook.com), but like traditional ads, widgets had increased success if the brand was relevant to the website. 
Background and Research Questions.
Although banner advertising still dominates the web, advertisers and publishers now have a much wider array of choices due to advances in interactive web technology and the rapid rise of social networking. Beyond merely clicking, users may now read special holiday or sports-related content that is “brought to you by” advertisers. Or they may create digital products, like customized cars or weekly food menus, and then either save them for themselves or give them to friends. Companies now routinely create profiles on social networking websites and encourage users to become fans or even display corporate logos on their personal profiles. And of course, traditional banner ads or email-newsletters still exist and are widely used.These ad types differ in the cognitive, social, and motivational foundations that make them a success, as well as the time and cost to bring them to market. As advertisers continue to transition marketing dollars to the web, and as publishers compete to offer effective inventory, empirical research is needed to make informed choices.
This study tested the key questions at right. Whereas other research might focus on marketing spending or performance metrics, this study concentrated on website users’ attitudes toward different ad types and the brands they promoted.
 To address the above questions, Allrecipes.com, the leading cooking/recipe website, commissioned Psychster Inc. to design a multivariate online experimental survey. Mockups were created of seven ad types (see descriptions below), appearing on one of two publisher websites (Allrecipes, or Facebook), and promoting one of two brands (a leading soup brand and a leading car brand), for a total of 28 combinations. To capture the interactive nature of the ads, a video was recorded showing the screen activity of someone interacting with the mockups, complete with a narration. For example, the narration for the sponsored content ad went, “The scenario is: you’re on [website]. You look at the current holiday page for St. Patrick’s Day. You notice that this page is sponsored by [brand]. You decide to look at Irish cookie recipes. You look over the recipe and again notice that this page is sponsored by [brand].”
These videos were imbedded in the survey, and respondents were randomly assigned to view only one. Afterward, they rated their attitudes on a variety of metrics. Designed as a multivariate experiment in this way, we were able to conclude that any difference in the metrics was in fact caused by the ad-type to which people were exposed.
 All surveys were fielded in May, 2009. The variations of the survey that depicted the ad types on Allrecipes were shown to 478 Allrecipes users who were recruited via a site-wide popover. The variations that depicted ad types on Facebook were shown to 681 Facebook users recruited via a panel company. Thus it is important to note that Allrecipes ads were only tested among Allrecipes users and Facebook ads were only tested among Facebook users. This decision was made to increase the ecological validity of the study by being more naturalistic, although it prevented us from testing the effect of the website independently of the website users.
 After viewing the video, respondents answered a series of questions both about how they viewed the brand featured in the video, and how they viewed the activity shown in the video. Within each group, there was a high degree of correlation between all the metrics, so only the most informative are reported here (see below).
Psychster Inc. 2010. Psychology of social media. http://www.psychster.com p4/6
Results. An analysis of variance (ANOVA) revealed significant differences across ad types; multivariate F(6,1065)=2.49, p < .001.
 1. Sponsored Content produced the highest interaction ratings, but the lowest purchase intent and viral recommendations. The green squares in the figures above show that people reported being more likely to interact with sponsored content than with any other ad type (ps < © Psychster Inc. 2010. Psychology of social media. http://www.psychster.com p5/6 .05). And they were least likely to see it as an ad. However, sponsored content was least likely to elicit purchase intent and a recommendation to friends. Thus, sponsored content was the most engaging of the ad types, but least likely to trigger consumer behavior or viral spread.
2. Corporate Profiles caused higher purchase intent when people could become a fan and put a logo on their own profile than when they could not. The blue circles in the figure above show that purchase intent was significantly higher when people could become a fan of a corporate profile than when this functionality was absent (p = .03). Similarly, people were significantly more willing to recommend profiles to friends when they could become a fan than when they could not (p = .03). If this were a field study, it would be unsurprising that customers who “brand themselves” with a company’s logo are more likely to buy from that company and recommend it to others. But because this was an experimental study where naïve users were randomly assigned to view such profiles, it suggests that this allowing people to become a fan and add a logo to their own profiles causes an increase in purchase intent and recommendations. 3. Give and Get Widgets were more engaging than Banners and Newsletters, but not more likely to increase viral recommendations or purchase intent. Like Profiles with Fans, Give and Get Widgets were the best all-around ad types across all metrics. People were significantly more likely to want to interact with ads where they could create and customize something (a custom car or a dinner menu) compared to banners or newsletters (p < .05). However, they were not significantly more likely to recommend the brands to friends or show an intent to purchase.  
 4. The above conclusions held across websites and brands.  The best and worst-performing ad types on the metrics as shown in the figures above were the same on both Allrecipes and Facebook, and for both the soup brand and the car brand. (The ANOVA produced no significant two-way interactions involving ad types.) Although future research should test additional brands and sites, there was no suggestion from this study that some ad types would only work for some brands in some contexts.  
5. No ad type was so engaging that it overcame the advantage found by matching the brand to the website. It is widely believed that that ads are at an advantage when the brand relates to the site on which it appears. Our findings replicated this effect, such that the soup brand performed better on Allrecipes than it did on Facebook. Despite the variability across ad types described above, no ad type was so engaging that it overcame this advantage. (The ANOVA produced no significant three-way interaction between brand, website, and ad type.)  


Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s