Copywriting Lessons From BuzzFeed's Worst Headlines [New Data]

As more and more people begin to publish online, more channels and content will be competing for your audience’s attention. A great headline is one of the most effective ways to break through the noise -- it has the potential to increase a post's traffic by 500%. So how do you write a headline that drives that kind of results for you?

Search for “how to write the best headline” and you’ll find many articles promising a blog title formula that will garner the most page views and shares. BuzzStream and Fractl decided to take a different approach: We looked at a set of low-performing headlines to identify key patterns and figure out what not to do when writing headlines.

To find this out, we turned to a publisher that dominates in social shares: BuzzFeed. Using BuzzSumo, we pulled all the English-language headlines from April 2014 to April 2015. We organized headlines by the first two words -- and any that began with a number received a value of “X” (e.g., "X Pieces," "X GIFs," etc.). The sample was narrowed down further by those that had at least 10 records, leaving us with more than 5,000 titles and their corresponding social shares.

Below is a summary of what we found.

Headlines with questions were some of the worst-performing types in the study.

"Are You" was one of the worst two-word combinations in a headline, garnering slightly more than 14,000 shares per article. It was representative of a larger pattern: Headlines that were posed as questions were some of the worst-performing of the study. Further analysis revealed:

• The "Are You" headline was one the lowest-performing for three social networks -- Facebook, Google+, and LinkedIn.
• Excluding the most popular sharing network (Facebook), "What Should" was also one of the least popular headlines.

After comparing all the question headlines to the 15 most-shared headlines, question headlines earned 24,951 shares per article while non-question headlines earned 83,475 -- more than a 70% increase.

Drawing attention to an article that comes from a less authoritative source (such as a community member) is associated with fewer shares.

User generated content can bring an audience closer to a brand, generate “free” content, and enhance loyalty. However, our research indicates that readers tend to share these articles less often. Titles with "Community Post" earned the least number of shares out of the entire data set, and compared to the highest-performing headline ("X Pictures"), "Community Post" earned more than 845% fewer shares. Additional findings revealed the following:

• Aside from earning the lowest number of shares overall, "Community Post" also earned the fewest Twitter shares -- fewer than 200 per article.
• "Community Post" headlines earned the least number of shares from LinkedIn and Google+ (less than 1% each) and the majority of their shares (96%) from Facebook.
• "Community Post" was also one of lowest-performing headline types for three social networks: Facebook, Google+, and Twitter.

Without Facebook shares, poorly performing headlines earned as little as 341 shares per article.

Facebook dominated social shares: The platform earned 85% or more of the shares for the 10 highest-performing headlines and 70% or more for the 10 lowest-performing headlines (the lower percentage stemming from two titles exceeding four-digit shares in both Facebook and Pinterest – "X Pieces" and "The Ultimate").

Without Facebook, 8 of the 10 lowest-performing headlines earned ~90% fewer shares. "What Should" also became the lowest-performing headline, earning only 341 shares per article. Additional findings revealed:

• "X Pieces" was affected the least: Excluding Facebook, the headline still earned more than 5,100 shares per article, and more than 25% of these shares came from Pinterest.
• 8 out of the 10 worst-performing headlines on Facebook also earned the fewest shares overall.
• The headline type with the strongest performance -- "X Pictures" -- earned more than 112,000 shares from Facebook. The lowest-performing headline type -- "Community Post" -- earned fewer than 12,000. That’s nearly 90% fewer shares.

Pinterest had the highest number of unique worst-performing headlines compared to other social networks.

Shares in Pinterest proved to be valuable for titles such as "X Pieces" and "The Ultimate," but our research also revealed that high shares from the network were harder to earn. Seven out of the ten lowest-performing headlines on Pinterest were unique to the platform, meaning that it had very few low-performing titles in common with other networks. Further analysis indicated:

• 3 out of the 10 lowest-performing headlines referenced males -- "A Man," "A Guy," and "This Guy."
• Pinterest was one of only two social networks on which a specific number was mentioned in a low-performing headline -- "The 15" on Pinterest and "The 25" on Twitter.
• Headlines with "A Man" earned the fewest shares on the social network -- only two per article.

It’s also interesting to note that using “guy” over “man” paid off: While "A Man" only earned two Pinterest shares per article, "A Guy" and "This Guy" earned 29 shares and 33 shares per article respectively -- up to a 1,550% increase.

Mentions of something specific (such as "whiskey" instead of "alcohol") tend to be common in articles with low shares.

One of the more interesting findings came through the analysis of word combinations. We looked at the top three-word combinations for the 10 lowest-performing headlines and the most popular two-word clusters in the bottom quartile. Our results revealed:

• References to two specific holidays -- Christmas and Valentine’s Day -- were found in the most frequent three-word combinations for "X GIFs" and "X Gifts" respectively.
• Although the Internet (and particularly BuzzFeed) can seem full of cats and celebrities, "Cats Who" and "Taylor Swift" were two of the most frequently used two-word clusters in headlines that earned low shares.

The biggest takeaway from our research? Headlines matter: The 10 lowest-performing headlines earned 415% fewer shares than the 10 highest-performing.

If you'd like to dive deeper into the research, you can check out the following infographic and review this deck, the latter which compares additional synonyms and looks at which months low-performing headlines are shared the least.

Study by Fractl and BuzzStream.