Methodology

The best practices examples from newsrooms were obtained through a series of phone conversations with engagement editors for primarily legacy print media organizations across the United States. These conversations discussed, among other topics, how these media organizations considered and approached voice in different posting scenarios, challenges they have overcome, challenges they still face and practices they’ve found work for their organization.

The content analysis, consisting of more than a thousand Facebook posts, was conducted using Facebook Insights data for three date ranges in 2015: Jan. 27 to 29, March 8 to 10 and May 21 to 23. Engagement editors at legacy newspaper media outlets were asked whether they would like to participate in the optional content analysis portion of this study following the initial conversation described above. The following 10 news organizations participated:

• Arkansas Democrat-Gazette
• The Beaufort Gazette
• The Courier-Journal
• The Island Packet
• Las Vegas Review-Journal
• Philadelphia Daily News
• The Philadelphia Inquirer
• Philly.com
• The State
• The Virginian-Pilot

Information from the following categories in Facebook Insights spreadsheets was used in the content analysis:

• When it was posted (date and time), permalink, post message and type, all found in the “Key Metrics” tab;
• Numbers of likes, comments and shares, all found in the “Lifetime Post Stories by Action” tab;
• And number of link clicks, found in the “Lifetime Post Consumers by Type” tab.

Each individual post was examined and subsequently coded for the following characteristics:

• Media organization (so that data could be returned to partners)
• Content topic: business, community/entertainment/lifestyle, education, government, health, public safety, social issues/religion, sports, weather, other (includes newsroom personnel posts, A1 photos, advertisements and sponsored content)
• Post voice: casual, direct story excerpt, formal, headline, no text
• Content voice: casual, formal, N/A (content that’s not text-based)
• Content type: feature, narrative commentary, news story, no text
• Post purpose: to seek information, to send some information and seek other information, to send information
• Content emotional response: controversial, negative, positive, weird/funny, other (no emotional response)
• Post language: first-person, second-person, third-person
• Post multimedia: Facebook page preview, graphic, photo, video, other (includes a shared Facebook event or group, or a text-only Facebook post)
• Types of post comments: discuss event/story, discuss coverage, discuss personal connection, respond to query, incite personal attacks, insert comment unrelated to post, none
• Original (versus third-party) content: yes, no

Posts were then compiled and examined in two different ways. What characteristics got the most engagement overall were determined using the entire data set by calculating average likes, comments, shares and link clicks for each individual characteristic, then comparing all of the characteristics in one category. The correlations made between types of engagement and the characteristics were made using data sets that excluded posts about weather, designated as a major outlier, in which posts received one or more of the engagement metric being measured (example: post likes = 1+). These statistically significant correlations were calculated through multiple linear regressions and Pearson correlations using Microsoft Excel and the plugin StatPlus. Subsets of data, when used, are specified throughout the study.