Date post: | 18-Aug-2015 |
Category: |
Data & Analytics |
Upload: | henriette-cramer |
View: | 112 times |
Download: | 2 times |
Effects of ad quality & content-relevance on perceived content stream qualityHenriette Cramer Yahoo Personalization & Platforms Sunnyvale, CA, USA
An illustration:
Ads that individually have been rated as high quality can have negative effects in context.
bad ad1. ad that is annoying to the user (Goldstein & McAfee, 2013)
(+ decreases product engagement)
content content
content
ad
contentad
contentcontent
content
content
Native advertising search ads, in-stream ads, sponsored posts
ad
content
content
content
content
Research shows …
Noticeable ads get more engagement (Goldfarb and Tucker, 2011)
Overly noticeable, annoying ads cost both user & site (Goldstein & Mcafee, 2013)
Relevance increases ad effectiveness (Goldfarb & Tucker 2013, Buscher et al., 2010, Yan et al. ’09)
Personal relevance can lead to more pleasant UX (De Sa et al, 2013)
Too personally relevant is too close for comfort (Agarwal et al., 2013)
Two-step studyMechanical Turk (US)
1.
ads rated as ‘good’ or ‘bad’
2.
Take good/bad un/related ads
Rate stream quality.
Step 1:45 native ads x 10 ratings per ad = 450 ratings 98 participants
Likert-type scales (1-5) + Why? for 1-3.
1. … how annoying would you say that this ad is? 2. … how would you rate the design of this ad? 3. … rate the trustworthiness of this ad? 4. how familiar are you with the brand advertised?
Step 2Ads in context
4 conditions,
Good Ads: 2 ads rated as not annoying & not untrustworthy
• 1. 2 good ads non-related to content
• 2. 2 good ads, 1 content-related: a credit card ad near credit-related headline.
• 3. ’Bad Ads: 2 bad ads rated as annoying & untrustworthy
• 4. No Ads: only articles
237 participantsLikert-type scales (1-7) + Why?
1. How credible do you think this news site is? 2. How would you rate the quality of this collection
of news links?
Limitations
1 ad format 1 type of content-relatedness 1 specific credit-related ad not looking at millions of actual users
… go forth and do a better study!
CC Steve Hersh flic.kr/p/7YqHge
should've, could've, would've
People distinguish between credibility of stories vs. overall site quality
Quality•No ads > both good and bad ads (MW, p<.01) •Good non-related > good content related (MW, p <.01
• High quality ads in context can have negative effects
• Interesting: Having no ads is better than good, but too content-related ads, but not sig different from good, unrelated ads
Step 2Quality Results
sig dig for credibility (H=9.083, df=3, p=.028) & quality (H=14.997, df=3, p=.002)
Mann-Whitney 1-tailed comparisons between conditions, with Holm-Bonferroni adjustment: target p value/significance n-rank of pair +1.
Branding goes both ways
“I think it's credible because [companyX] is one of the advertisers and [companyX] is a very reputable company…”-GCR
Ads were mentioned in ‘why’ people perceived the stream as high or low quality
whether they were there or not.
Two positives can be a negative
Extends Goldfarb & Tucker, 2013:
…combining multiple effective ad strategies can decrease their effectiveness when each alone prime user concerns.
Ads that individually have been rated as high quality can have negative effects in context.
Implications for ad-supported services:• Clearly distinguish ads & content • Evaluate ads for quality, then evaluate for
quality within-context • Offer in-context feedback opportunities
Ads affect experiences, research them!
henriettecramer.com @hsmcramer