How News Images Affect Clicking on Subscription Appeals

ABSTRACT

The newspaper industry’s business model increasingly relies on subscriptions as a means of revenue. Little scholarly research has examined what characteristics of subscription appeals make them more or less successful, although prior research on news attention proposes that negative images may be more successful whereas digital advertising research suggests the benefits of positive images. In 10 tests with three newsroom partners, we experimentally manipulate the type of image used in the subscription appeal. Across the tests, we varied the type of subscription (paid or free newsletter), the medium used (Facebook Sponsored Posts or direct email), and the targeting method (Facebook custom audiences or email lists from the newsroom). Results show that images of journalists doing their work or images of natural disasters covered by the news organization are consistently more effective than logos at generating subscriptions. Implications of this study exist for our understanding of the effects of visuals, for news organizations attempting to garner subscribers, and for researchers interested in conducting real-world experiments via Facebook or email.

KEYWORDS:

• Email
• experiment
• Facebook
• images
• news
• subscriptions

Acknowledgments

We are grateful for the time and dedication of the three newsroom partners who helped with the experimental portions of this research and the staff from the newsrooms who shared their insights on gaining donors, members, and subscribers. We also thank the entire Center for Media Engagement team, especially Alishan Alibahi, Cameron Lang, Melody Avant, and Callie Reynolds.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 We use images of journalists instead of images of news subjects purposefully. Newsrooms may be reluctant to use images of people suffering in their advertising for ethical reasons; they want to avoid seeming as though they are looking to profit from others’ distress. Indeed, we heard this reaction in the course of the research reported here.

2 For the Facebook campaigns, the optimization and features related to placement for each newsroom were held constant. All ads ran for a full 14-day period, instead of ending the campaign when a “winning” ad was found. Each campaign was optimized for traffic to a website (either the newsletter signup page or online subscription page for each newsroom). The only placement used for these ads was the Facebook News Feed. Each Facebook ad was also optimized for link clicks and paid for by impression. The daily spend rate for each campaign was set at $85.00 (the large regional newsroom’s daily spend was split across the two contact methods). The call to action for each ad was “Learn More.”

3 The minimum number of studies required for conducting a meta-analysis is two (Valentine, Pigott, and Rothstein 2010, 241).

4 We use random-effects models because we did not believe, a priori, that each study would provide an estimate of a common treatment effect. In this situation, random-effects models are recommended (Chen and Peace 2013, 117–8).

5 Since we cannot observe the true effect sizes in the random-effects model directly, the estimated true effect size (${\hat{\tau }}^2$) and its variance are calculated from the observed data using the DerSimonian-Laird method (Chen and Peace 2013, 115).

6 Higgins et al. (2003) suggest a categorization of the $I^2$ index as low, moderate, and high to 25%, 50%, and 75%, respectively. The higher value of the index, the greater inconsistency among subgroups. $I^2$ is considered as statistically significant if the lower limit is greater than zero.

Additional information

Funding

This work was supported by grant funding from the American Press Institute to the Center for Media Engagement.