Cracking The Code On The Web’s Most Shareable Images

A startup called Curalate is developing a product that predicts which images will have the most success on platforms such as Instagram and Pinterest. Brand images without faces and red pins get more repins, for example.

Cracking The Code On The Web’s Most Shareable Images

With the right tools, brands can easily monitor words, such as their names, across social platforms. But increasingly consumers aren’t sharing words. They’re sharing photos.


“Consumers are having millions a conversation a day, saying, ‘I love this J.Crew sweater,’ without saying the words ‘J.Crew’ or putting in a SKU number or doing anything that says, ‘This is a J.Crew sweater,’” says Apu Gupta about images shared on networks such as Pinterest.

Two years ago, he cofounded a company called Curalate that shows brands how their images are shared on Pinterest and Instagram–monitoring, for instance, the top products being shared from a brand’s website and the traffic coming from images posted on social sites. Brands use these insights about their past performance to improve their current performance.

Now Curalate hopes to skip a step by predicting which images will work best on a brand’s profiles before they post them. To do so, the startup is studying what makes an image succeed on social networks, and it recently it analyzed 32 different visual characteristics of 500,000 Pinterest images.

Here are some of the resulting observations about what type of images get shared the most:

  • The most repinned images have multiple colors.
  • Very light and very dark images are not repinned as often.
  • Completely desaturated images (gray) and completely saturated images have fewer repins than images that are more moderately saturated.
  • Red pins get repinned more often than blue pins.
  • Images with less background space get repinned more often.
  • Brand images without faces receive more repins.
  • Images with a smoother texture are up to 17 times more repinned than images with a rough texture.

None of these observations are quite as succinct as “Facebook posts that include the phrase, ‘Like this,’ get Liked more often.” In most cases, software would be needed to help decide which images fit within these guidelines. It would also be possible for the software to learn which types of images a specific brand’s customers like best.

“In theory,” Gupta says, “a brand could take an image, drop it into our platform, and our platform could score it [in order to decide which image to use].”


But isn’t it possible that I shared an image not because it’s a colorful, moderately saturated, red-toned, full-frame shot without any faces or rough textures, but because I really like the pair of earrings it depicts? Gupta says the large sample size controls for that. “In the question of ‘Oh, I like these earrings,'” he says, “still there’s a question of ‘Why did you pick that image of the earrings?'”

[Image: Flickr user Theilr]

About the author

Sarah Kessler is a senior writer at Fast Company, where she writes about the on-demand/gig/sharing "economies" and the future of work.