An interesting MIT study titled, What makes an Image Popular? helps us reveal what is it exactly behind an image that makes it more popular than others on social media. They have used Flickr as a base to their stydy and a data set of 2.3 million images. The study focused on the impact of social cues, the content of images and their impact on the ‘popularity’ of an image online. Many of these conclusions are not surprising, but here I will give a short overview of the study. The study itself can be found on the link here: MIT study.
The Impact of Colour
More striking colours lead to more popular images. Hues of green, blue and grey are less popular and they are images that have not been viewd a lot on Flickr. But bright shades including aqua, red, navy and pink correlate with more popular images.
The Impact of Objects
The press coverage that followed the study’s publication undeniably focussed on one thing: sex sells in images. The study created 4 categories and here are some examples of what items are more popular than the others.
- Strong positive impact: miniskirt, maillot, bikini, cup, brassiere, perfume, revolver
- Medium positive impact: cheetah, giant panda, basketball, llama, plow, ladybird
- Low positive impact: wild boar, solar dish, horse cart, guacamole, catamaran
- Negative impact: spatula, plunger, laptop, golf cart, space heater
Bikinis, bras, revolvers and miniskirts are visual cues for a popular image; spatulas, plungers and golf carts are on the other hand are not (pretty obvious too!). Sex doesn’t necessarily sell, but it definitely drives photo views on Flickr. Here it is definitely worth to consider is "a view" what we are after. Our goals might be share or like or comment. Sex might not be "the thing" people necessarily openly like and comment, whereas they might "sneak a peak", so to say. So it is easy to suspect if sex on photo is actually helping you.
The Impact of People
This study has not discussed people on the photos. How do people on the images influence the popularity. There is another study that helps us here: a Georgia Tech study of 1.1 million Instagram photos found that any image containing a human face is 38% more likely than any other visual subject matter to be liked and 32% more likely to attract comments. An interesting step forward could be a study of details in those people on the images. I guess we can here do our own social experiments with our own photos using Photofeeler.
The study also came out with a a basic tool that claims to be able to predict how popular an image will be online. The tool is pretty simplistic, but it can be handy when you are wishing to compare 2 images.
Overall it is an interesting read of course but take into account some of the limitations here - is "views on Flickr" really the best base-line measure?