Twitter best practices: A comparative study of regional media

Kurt Gessler
Mar 7, 2016 · 7 min read

There’s a certain sameness that has pervaded the main Twitter accounts of large U.S. media outlets as the platform has matured. In fact, during major breaking news, it’s often an echo chamber of essentially the same tweet, usually with the same photo.

So when I set out to quickly determine how unique we, at the Chicago Tribune, were in operating @chicagotribune, I wasn’t sure what I’d find.

I mean, we know the accounts that abuse all-caps intros or layer random hashtags.

Say, did an ancient bridge collapse?

But could there be a wider array of nucleotides in the genetic makeup of a primary brand account? Turns out the answer is yes, and it’s far beyond just content. Part of it, I suspect, has to do with Twitter itself. Unlike Facebook, where posting too often or posting the same link multiple times can limit your reach, Twitter lacks that punitive algorithm. This has allowed some freedom in approaches. And you see that freedom when you drill down.

Before we start, I should answer the one obvious question: Is Twitter still worth it? To us, yes. In 2015, Twitter (t.co) was a top 5 referring domain, sending more visits to Chicago Tribune than Bing and Reddit combined. Sure, it’s not Facebook nor Google, but it’s meaningful.

To begin, I had to determine a pool of accounts to study. I wanted not only a similarly sized Twitter following, but also similarities in newsroom mission and focus. Then I wanted to pick a few more progressive examples that might, perhaps, be aspirational. Given those guidelines, I used Alexa, Pew and Nielsen data and settled on these sites:

My baseline accounts had similar web traffic and markets. That sample corresponded to these Twitter accounts:

My initial goal was to take a quick cut of weekend and weekday tweets, looking at frequency, content, native media, redundancy and RTs. And hopefully, through that study, I could find some areas to improve our process.

I went in looking at these areas:

  • Total tweets for a day (24 hours)
  • Total w/media
  • Brand RTs
  • Personal RTs
  • Outside account RTs
  • Live tweets during ongoing event
  • Redundant tweets (tweeting the same story on that same day without any updates)

I also grouped content by section/category: news, sports, real estate, business, entertainment, commentary/opinion, lifestyle, live stream, promotional (as in self-promotional), history

Lastly, I looked at what native media was included: photo, photo gallery, gif, Vine, video, Periscope, poll, map/graphic/illustration, quote card

Immediately, it was clear those generic newspaper sections would be insufficient. So, for transparency, here’s how I grouped the following under those larger sections:

  • News: local, national, international, weather, politics, policy, news features, obits, education, science, transit/transportation, studies/reports
  • Business: tech, internet/social, jobs, auto, commercial development
  • Life: dining, travel, general health, parenting, pets, books, art, fashion, dating/relationship
  • Entertainment: celeb, theater, video games, music, TV, movies, ent obits, fine arts

Another aspect that became immediately clear was the enormity of the task. While some tweets could be easily categorized, many couldn’t. Even the full URL often couldn’t help if one site put celebrity news under a lifestyle path. This was wickedly manual, involving clicking through each link and weighing the content.

Another challenge appeared as I started to look at Vice and Vox. It simply was not always easy to categorize content. On Vox, it was problematic to separate political reporting from commentary, first-person opinion or observation. And if it aggregated shows like Samantha Bee or John Oliver, are those entertainment or commentary? And Vice was worse with tags like Views my Own. Where would you categorize these tweets?

What Hitler’s supposedly tiny, deformed dick tells us about how we view evil

Here’s how to start fights and insult strangers in nine foreign languages

Photographing the gender fluid tribes of the Amazon

Despite these issues (Hitler’s genitalia I labeled commentary), I took weekday and weekend samples from all the sites, bucketed the results and found some interesting data that will change how we approach @chicagotribune (raw data in Google sheet).

Frequency:

Across the sample, the average number of weekday tweets was 95, with 59 on weekends. Boston, by far, had the highest average, with 164 per weekday. Weekends were generally lower, especially at Houston, which had only 10 on the one weekend day I looked.

Non-traditional media had fewer tweets, but really didn’t stand out in structure or procedure. Content set them apart more than style.

Content:

The Washington Post was, by far, the most news-centric Twitter account in our sample. Vox finished second. Generally, all the regional media outlets favored news-related content.

The spreadsheet, after that, really shows each media site favors certain sections, like sports. Few, however, hit commentary and opinion very hard.

Redundancy:

Here was a surprisingly disparate segment, where the Boston Globe was once again atop the list. To be clear, this is where I could infer that a piece of content was tweeted another time without that story being updated.

The Globe, in fact, often tweeted the exact same story a half dozen times in a single day. To its credit, however, its staff often disguised it well. Here’s one series of examples:

Honestly, this was no social media accident. That team was working hard to sell what they must perceive as high-value content to hopefully a new audience that missed it the first time.

Here’s Vox doing a worse job.

Only 1 of these 6 stories hadn’t been tweeted previously, although all of these occurred between midnight and 5:15 a.m., so it’s not exactly prime time. Still, it feels more like scheduled tweets to hit a quota.

Native media:

This was an important area to look at because, by Twitter’s own study, photos boost engagement 35% and video 28%. And that was in 2014. So we know you need more than just text to succeed at Twitter.

These results genuinely shocked me. The new media types fell far below legacy media like the Dallas Morning News, Tribune and Globe. And Dallas pretty much made sure every item had something complementing the words. But the successful and progressive Washington Post was pretty much just words.

If you dive more deeply, Dallas, Houston and Boston also did a good job sprinkling in gifs and video while the Tribune was too reliant on just images — although no one used more native photo galleries.

Personal or sub-brand accounts

This was another area that was wildly inconsistent. Some accounts RT’ed their sports or entertainment accounts a lot. Others never. Some relied heavily on reporters and editors and RT’d them. Vice and Boston did not.

Given that either method limits the amount of data you have in Twitter Analytics, I would in hindsight urge moderation to both. Plus, those tweets have a lower chance for someone to follow your primary account. The Tribune is clearly too high on brand and Houston and Dallas and too high on RT’ing personal accounts. I mean, how good a UX is this series from Detroit?

Live streams

As they had them available, most accounts promoted live video. I did think the Globe one-upped everyone by live tweeting quotes and re-promoting the stream.

In summary

First and foremost, this sample size is too shallow to make sweeping declarations. Because this was a manual process, there’s limits to what could be effectively culled and categorized. I would love to develop a tool to programmatically track all these categories and sections and more. I’ve used Social Report and Simply Measured but neither do all this. And beyond just more data, the end game has to involve correlating this data to engagement and growth in followers. This is what they do and this is the result. That’s where this becomes the most valuable. Still, despite the sample-size limitation, it was valuable.

What will I take away from this for the Tribune?

  • Far fewer brand RTs; maybe a few more personal account RTs
  • More tweets with native video, gifs, maps, polls—just more variation in native media
  • Hit good topics a bit harder, but make each tweet unique
  • Our frequency is at or above with similar sites. We don’t need to tweet more
  • Our content category balance is par with similar sites. But there might be an opportunity to promote more opinion/commentary
  • Find more engaging ways to promote live streams

I’ll likely share our best practices on Twitter in a coming post, guidelines which will incorporate these findings.

Kurt Gessler

Written by

Director of Editorial Ops at @ChicagoTribune et al. I also teach stuff at @UNLincoln @Unl_CoJMC. Practicing journalism et sic per gradus ad ima tenditur