Finding a voice on social media: Insights for local newsrooms

As more people, especially in younger generations, rely on social media to stay informed, news organizations must figure out how to speak to readers in these personal environments. How can a journalist’s news posts stand out in a newsfeed of both current events and family vacation photos?

What we found by examining this question is that there’s not a simple answer, or even a single one. Several factors about how a social media post is written affect the engagement it receives. They include the number and demographics of account followers, time of posts, ever-changing platform algorithms and voice, defined for the purpose of this study as how a social media post might sound when read aloud.

As part of this year’s American Press Institute summer fellowship, I gathered insights about the impact of voice on social media engagement in two ways. I talked with engagement editors across the country about best practices in their newsrooms, challenges they’ve overcome and challenges they still face. I also partnered with 10 local newsrooms and performed content analysis of more than 1,200 of their Facebook posts to gather quantitative insights about what resonates with users — voice and other factors.

[Apply now to be the American Press Institute Summer Fellow for 2016]

I found that voice varies among media organizations. Some engagement editors said they are afraid to create a unique voice; others are struggling to find one. Those who have defined a voice and developed a strategy might strive to come off as an authoritative source for the community, or as a member of the community itself — more like a know-it-all neighbor who can be a bit sassy.

We identified six basic types of “voice” that news organizations and journalists use for Facebook posts, some more successful than others:

1. Just the link

Posts can just include a link with no other identifying information; this tends to happen on Facebook, where inserting a link causes a box with other identifying information to appear.

Just 2 percent of posts in the content analysis were like this; the most common instances were sharing another user’s or page’s photo or sharing major community news. With the exception of a few posts about major community news, engagement on these types of posts was typically low.

2. Just the headline + link

Posts can consist of just the headline and the article link. Sometimes, the person posting might swap out a word or two — ”local” for “Atlanta,” for example — to add information that might be obvious on the news organization’s website but out of context on a feed of international information.

Just 4 percent of posts in the content analysis only included a headline with the link. Headlines for hard, breaking news seemed to drive decent site traffic when looking at link clicks overall. But many newsrooms say that long-term — and across all content — posts with headlines don’t drive as much engagement as posts that included other text.

“When (producers) take the time to really punch up their writing, we see a lot better results than if we just run a headline,” said Brian Smith, The Des Moines Register’s engagement editor.

3. Excerpt from the article

Posts can be direct excerpts from the text of the article. These excerpts are often ledes, nut graphs or interesting quotes or statistics. They can be, but aren’t always, enclosed in quotation marks.

About one-fourth of the posts analyzed contained these direct excerpts — and of those posts, about three-fourths contained excerpts with a formal news-writing voice, and about one-fourth with a casual voice. This style was most common for subjects that tend to be more serious and complex, such as education (about 33 percent), government (about 34 percent) and health (about 38 percent).

When just looking at posts that contained direct story excerpts, content shared that was not text-based — specifically, Storifys — received more link clicks on average. However, there were only a few Storifys included in the sample.

4. Blurb with formal news writing

Posts can contain polished language that one might use in formal, published writing — language that wouldn’t be out of place in a traditional newspaper article. These posts often use complete sentences to share an interesting fact or reword the nut graph. If these posts use puns, they’re sharp and witty.

About one-fourth of the posts analyzed used this original, formal voice. In the content analysis, there were a higher proportion of formal posts for topics that tend to be more serious; for example, about 43 percent of posts about public safety (which includes crimes, courts and traffic accidents) are written like this.

When looking just at posts using a formal voice, more serious topics performed the best. Posts that shared hard news content received more likes, comments and shares on average. Posts sharing content likely to evoke a negative emotional response received more comments, shares and link clicks on average. And posts sharing content likely to evoke a controversial emotional response received fewer likes but more comments, shares and link clicks on average.

5. Blurb with conversational language

Posts can be casual bursts of information conveyed in a way you might speak to a friend or colleague. These posts could come from a specific persona, but they’re often just conversational in nature. These posts might be sassy or snarky, use cultural references, include slang or abbreviations, or make a joke or pun. They could also include a non-controversial truth or opinion, such as how this dog might be the cutest thing you’ll see all week. About half of these posts include first- or second-person language, much more than any other type of post, content analysis shows.

About two-fifths (43 percent) of the posts analyzed used an original, casual voice. The content analysis found that topics with lighter content — such as those about community, entertainment, lifestyle, sports and weather — had more casual posts than the average topic (about 48 percent).

When looking just at posts with a casual voice, no one characteristic stood out as bringing in more likes, comments, shares or link clicks on average. However, posts that shared photos received fewer likes on average, and posts sharing content that elicited a positive emotional response received fewer comments on average.

6. The cliffhanger

Posts can use suspenseful language and leave out key bits of information to mislead people and entice them to click on the story link, hence the term “clickbait.” None of the posts we looked at were classified as this.

Many engagement editors want to stay far away from this style Upworthy and others popularized (and that Upworthy is even turning away from now). “I try to be really careful about not posting something that I can’t deliver,” said Irene McKisson of the Arizona Daily Star. “You lose trust that way really fast.”

What content analysis showed

Looking at more than 1,200 Facebook posts from 10 newsrooms, we found that although newsrooms employed various tactics when writing posts, most newsrooms tended to write a majority of their posts in one consistent voice.

We looked at more than voice, however; we also looked at other characteristics of posts and the content shared in them. Below are the characteristics we looked at and, overall, which characteristics received the most engagement on average across newsrooms:

• The voice of the post, as we detailed above. Posts with a casual voice, a direct story excerpt or no written text got the most likes on average; posts with a casual voice or no written text got the most comments and shares on average; and posts using a formal voice or with no text got the most link clicks on average. The few posts without any text either had very little engagement or, if about major community news, had a lot of engagement, so this is likely an outlier when making recommendations for posts in general.
• The type of post shared. Posts that shared an external link and posts that shared two or more photos got the most likes, comments, shares and link clicks on average. However, most of the posts in this sample were links, so further analysis with a more diverse sample could be beneficial.
• The purpose of the post. Posts that sent some information and sought other information — for example, shared information about a news story and asked followers a question — received on average more likes, comments and shares than posts just performing one of those functions. However, posts that just sent information about a topic received the most link clicks on average.
• The language used in the post. Posts written in second-person got the most likes and shares on average; posts without any text or written in second-person got the most comments on average; and posts written in second- or third-person or without any text got the most link clicks on average. The few posts without any text either had very little engagement or, if about major community news, had a lot of engagement, so this is likely an outlier when making recommendations for posts in general.
• The multimedia shared in a post. Posts with an embedded photo, including those in the link box, got the most likes, comments, shares and link clicks.
• The topic of the news reported by the post. On average, posts about weather, business and community, entertainment or lifestyle got the most likes; posts about weather, business and public safety got the most comments; posts about weather and public safety got the most shares; and posts about public safety, weather and business got the most link clicks. Posts sharing content about weather proved to be a significant outlier; content about weather received on average many more likes, shares, comments and link clicks than other content. We excluded these posts, which would skew the data set in a statistical analysis, for the overall correlations discussed later to get a clearer picture of what works.
• The voice of the content shared. Posts sharing content written in a formal voice saw slightly more engagement on average than posts sharing content written in a casual voice, but both of these voices performed a lot better across all metrics than content shared that wasn’t in a traditional article format — for example, just a video.
• The style of content shared in a post. News stories and features got the most likes, shares and link clicks on average, and news stories got the most comments on average.
• The likely emotional response a reader would have to the content shared in the post. Positive and weird/funny content got the most likes; controversial, negative and weird/funny content got the most comments; and negative and weird/funny content got the most shares and link clicks.

Our content analysis also looked at correlations between types of engagement received — likes, comments, shares and link clicks — and the characteristics of both the social media posts themselves and the news content shared in these posts. These provide an idea of what post characteristics have worked well for the legacy media outlets included in the analysis, a starting ground for characteristics that also might be successful in other newsrooms.

• Having a lot of one type of engagement means that, on average, other engagement metrics are higher, too. For example, the more likes a post has, the more comments there are on average. This applies to likes, comments, shares and link clicks.
• More likes on average went to posts that used a casual voice, included a direct story excerpt, sent some information to people and sought other information from them, or posts without any text. The few posts without any text either had very little engagement or, if about major community news, had a lot of engagement, so this is likely an outlier when making recommendations for posts in general.
• More shares on average went to posts that send information (as opposed to crowdsourcing) or that embedded a video.
• Fewer shares on average went to posts sharing photos or content likely to evoke a controversial, a positive, or no emotional response.
• More link clicks on average went to posts that shared news content or that just shared a link to an article. Again, the posts not including text could be an outlier.
• Fewer link clicks on average went to posts sharing content likely to evoke a controversial, a positive, or no emotional response.

More details about the content analysis will be discussed throughout this strategy study.

What’s ahead

Decisions about voice don’t end with scheduling a few social media posts. The same posture transcends how media organizations interact with their online community through comments, crowdsourcing and one-on-one interactions.

This study showcases examples of how media organizations, particularly legacy ones, have successfully developed and implemented a voice into content and interactions on social media.

Many of these examples are supported by content analysis of legacy newspapers’ Facebook posts, the findings of which will be explored throughout this best practices guide. It also includes suggestions for how to make social voice a priority in your newsroom — whether with one social media editor or a team of 10 web producers — and integrating it into your daily workflow.

The examples in the following chapters provide recommendations for voice best practices, as well as methods and resources that can be used to help integrate them into your newsroom:

• Developing a voice
• Implementing a voice
• Integrating voice into your workflow
• Interacting with your community
• Methodology

[newsletter title=”Don’t miss other API research and insights” blurb=”Get the latest from our team at API, along with our collection of fresh useful insights on the web each morning in our Need to Know newsletter. Sign up to stay in the loop.”]