How BuzzFeed, R29, and Other Top Publishers Optimize Their Headlines and Images

As Drake proffered in his hit single “Headlines”: “They know that the real is on the rise. F*ck them other guys. I even gave them a chance to decide. Now it’s something they know. They know, they know, they know.”

Wouldn’t it be great if all content marketers had the same confidence about their headlines and images as Drake does about his career? Well, they can—through optimization and A/B testing their content. Just give your audience a chance to decide on which variation they like best until it’s something you know, you know, you know. Okay, I’m done referencing Drake lyrics.

Upworthy curator Adam Mordecai said at Netroots Nation that you must test headlines because, when it comes to producing potentially viral content, you’re dumb. We’re all dumb, and our intuition often means squat.

Yet Webaholic reports that, according to Upworthy, a good headline can be the difference between 1,000 or 1,000,000 people reading your story. And optimizing images can do wonders for your SEO. Also, after testing through the best images and headlines for your website, you can use the information you’ve gleaned and apply it to strategies for email, social media, and future content.

Isn’t a 20, 50, or even 500 percent difference in engagement worth testing for? I think we all know the answer to that one. For brands still getting started with optimization, here’s what you need to know.

Why you should optimize headlines and images

Ky Harlin, lead data scientist at BuzzFeed, believes it’s crucial to put resources into creating strong headlines and images, because they are often the two things that introduce readers to content.

“It’s basically the only way we control the presentation of a given piece of content,” he says about optimization. “In order for someone to peruse an article on BuzzFeed, they have to first click through a headline or thumbnail on our site or on social media or somewhere else. So that’s our way to draw people in.”

Neha Gandhi, VP of editorial strategy for Refinery29, also stresses how optimization is important for making a solid first impression on readers.

“Headline testing is really the first step when it comes to figuring out the right approach to how we want to position a story to our readers,” she says. “We want to position [our content] in a way that’s the most accurate, the most honest, and really makes it clear what story we’re telling from the outset.”

It’s not just the clicks that publishers are chasing. It’s also, more importantly, the share, which is a metric that is at the core of BuzzFeed’s values. In fact, social sharing is the main driver of BuzzFeed traffic, which was estimated in 2014 to be about 150 million unique visitors per month. In order to drive up that share count, BuzzFeed’s editors have to make sure that they are offering headlines that truly represent the content that it teases to readers.

“You can usually get somebody to click on something just based on their own curiosity or something like that, but it doesn’t mean that they’re actually going to end up liking the content,” Harlin says. “So we tried to marry that idea with what we do, in terms of optimizing headlines and thumbnails.”

The headline is such an important part of a piece of content that you can’t afford to leave its success to chance. For example, an infographic by Marketo shows how Upworthy managed to draw 17 million views to a YouTube video that originally only had 1 million views. The difference? The headline.

The Iowa House Democrats’ YouTube channel titled its video “Zach Wahls Speaks About Family,” while Upworthy chose the title “Two Lesbians Raised a Baby and This Is What They Got.”

Those 16 million views are a big payoff for the simple exercise of testing out, and then optimizing, your headlines.

How to optimize headlines and images

According to Upworthy’s blog, Upworthy curators generate 25 headlines for every piece of content they want to post. The content team them narrows this list down to a few final contenders to be tested against each other. As noted in Upworthy’s SlideShare “10 Ways to Win the Internets,” some of your headlines will be really bad, and that’s okay. Headline #24 could be terrible, but if #25 is a gift from the headline gods, then it’s all worth it.

As mentioned above, images can also be optimized for site engagement and search traffic performance. Quick Sprout’s blog offers tips for making your images search-engine-friendly, including using alt tags to describe your image, using straightforward and descriptive file names, adding various types of metadata to your image file, and using rich snippets to tag an image to your website or business.

When it comes to headlines, Refinery29 uses content marketing platform Taboola (which just raised $117 million) to help test the performance of various options. There’s often a visible distinction between the results.

“Sometimes you can see as much as a 5 or 6 percent CTR lift, which makes a pretty big difference,” Gandhi says, “and it really tells us what our readers are responding to and what they want more of.” Gandhi and her team then apply the lessons they learn to how the content appears on their site, and even on their social feeds.

Publishers can also take advantage of leading optimization platforms such as Optimizely and OpBandit to A/B test on-site. These services allow you to plug in different headlines and images for a piece of content, as well as track engagements, clicks, and conversions in order to see which options your audience responds to best.Some companies, like BuzzFeed, have custom, built-in systems for their editors to use.

In BuzzFeed’s case, editors can input several headlines and images for a piece. For the first couple of hours after it’s published, visitors to the homepage or the article page will randomly get one of those variations. Then, editors test the performances against each other, taking into account click-through rates and share rates.

“Usually, one or two of the different variations outperform everything else,” Harlin says, “and then that’s ultimately what’s used going forward for the rest of time.”

How to optimize for email marketing

Testing headlines and images isn’t just good for your clicks. Optimization can also improve your email strategy.

For example, Refinery29 uses Sailthru to create different segments within their emails for stories. In order to make sure they’re sending out the best version of their email, they use headline and image testing as a way to optimize every email.

“Today we used a story about warm weather travel destinations … We picked two or three image options from inside the slideshow, and tried them out as the [featured image],” says Gandhi. “Upon testing, we saw that one of the images had a pretty big list [of clicks], so we swapped it in. We’re going to test it as an email later this week.”

According to Harlin, BuzzFeed’s primary goal for email marketing is to make the experience more of a personal one for the reader, since everyone on their email list has already opted in to BuzzFeed’s content. He notes that the website strategy has been more global, whereas the strategy for email is for user-centric. And the information that they can glean from those emails is vital.

“We can get quite a lot of data back from these users,” Harlin says. “As we start sending emails and also, as we see when they signed up, we can see which specific articles and which categories of articles they tended to like.”

How to optimize for social media

Once headlines and images are optimized on a publisher’s website, they can then take that information and apply it to social content. The best headlines often make the best tweets or Facebook posts, driving the ultimate volume of engagement.

”We often test stories that we think are going to be big on social, if we feel like they can wait,” Gandhi says. “We tend to test a few different ideas on how we want to position the headline or the call-out on social, and use that as the first testing ground.”

Publishers can also test out different headlines on Twitter or Tweetdeck, where it’s easy to schedule various tweets and track engagement on the platform, as clicks through link shortening platforms like

At BuzzFeed, Harlin says, “In terms of what we’re doing in the site with A/B testing… and best practices around creating headlines and thumbnails, I think that stuff pretty much directly applies to something like Facebook.” He notes, however, that there isn’t a whole lot of optimizing to be done with the Facebook posts themselves, as Facebook doesn’t allow for that sort of A/B testing on their platform.

“We have data from Facebook about how different presentations of the same story perform against each other,” he says, but it’s more used for handing over to the editors to “see over the hundreds of posts that they’ve created over the years and get a sense of what’s working and what isn’t.”

Everything BuzzFeed has done so far on Twitter and Facebook has been with the goal of driving traffic back to But some emerging social platforms like Instagram, Vine, and Snapchat, aren’t conducive for sharing URLs and linking back to a homepage. Instead, their goals are to generate engagement native to the platform itself, with tags, likes, and comments. Facebook and Twitter are also pushing for the same native platform focus.

For testing content on these platforms, BuzzFeed launched a new, experimental division called BFF. In terms of optimization, Harlin says they’re hoping whatever data they find from BFF content can be leveraged for optimizing content on the site, and vice versa.

How to optimize for future content

By testing different headlines and images, publishers can learn what readers respond to, and in turn inform decisions about future content. For example, BuzzFeed found that color images perform much better than black and white images, because in color images, it’s easier to discern what you’re seeing. The information is just relayed quicker.

They also found that images in which you can see someone’s entire body perform worse than the ones where you can just see a face or upper body. Again, this is probably because it’s easier to tell who or what you’re looking at if it’s close up.

In terms of headlines, Harlin says that shorter works better. This often has to do with how the story is displayed.

“On BuzzFeed, we have this thumb-strip at the top of every single page with a bunch of other recommended stories,” says Harlin. “You can hover over it, and it shows you a headline, but it can only show a certain amount of the text. For some super-long headlines, you can’t actually read the whole thing, so you don’t know what the story’s about.”

On a regular basis, BuzzFeed will have someone from the data science team talk to their editorial teams and sift through various insights they’ve collected.

“We’ll have a conversation with them and oftentimes, they want us to look deeper into the hunches that they have to confirm them.” And the cycle continues, with editors informing the data scientists, and the data scientists reporting back, all the while testing different pieces of content and acquiring more information about reader interests.

While intuition about what might perform well can help you start banging out that list of 25 headlines, the best way to be sure of what your readers want more of is by gathering data through optimization strategies. Only then can you reach Drizzy’s level of confidence.

Image by Shutterstock

Get better at your job right now.

Read our weekly newsletter to master content marketing. It’s made for marketers, creators, and everyone in between.

Trending stories