• Woman at the art gallery

Use AI to identify high-converting marketing images

Cutting-edge AI (artificial intelligence) research into how computers see marketing images is changing the way marketers build more successful campaigns.

Powered by machine learning algorithms, marketing technology tools now serve as behind-the-scenes assistants which empower you to perform visual analysis of marketing assets and get insight on what customers would prefer ahead of sending out their campaigns.

Marketers also face uncertainty when it comes to creating a digital marketing campaign, especially targeting new customer groups. We humans can’t tell which images will be appreciated by our audiences and what content they would like.

The perfect image can be the difference between a marketing campaign that’s memorable, engaging and successful, and one which just fails to capture the imagination.

Marketers are also biased. Sometimes we think that certain marketing pictures look good, yet they end up not delivering great performance. So how to ensure that the choice of marketing images is objective and will deliver results?

Artificial intelligence can help with this challenging task. Machines can now predict what visual assets will be appreciated by your audience, justifying their choice and educating you about things to remember for maximum engagement on social media and other digital channels.  

The art of selecting marketing images meets computer science

The perfect image can be the difference between a marketing campaign that’s memorable, engaging and successful, and one which just fails to capture the imagination.

Knowing what’s going to work best sits in the core of every marketers job. This is why many consider their audiences engagement and appreciation a badge of honour and mark of skill. Given that images impact your marketing dramatically, content creators should be very selective with what assets they rely on. The more they back their decisions informed by data, the more they can be sure about the future success of their campaigns.

Imagine the star stock traders in the financial world, who are now fully equipped with the most cutting-edge AI algorithms and tools. In this sector technological progress allowed moving from gut-feeling prognosticators towards data-driven decision-making.

Will these new technological forces such as Artificial Intelligence get marketers onto the same journey?

In many ways, the need has never been greater. The average customer is no longer leafing past hand-drawn adverts in the newspaper, but spending less than a second scrolling through countless calls to action on a screen about the size of your palm.

In this article, we’ll explore the intersection where marketing art meets computer science and how new technology is changing the way we build engaging visual campaigns.

What is computer vision and is it useful in marketing?

Computer vision is a scientific field of computer science that deals with enabling computers to understand images in the same way that human vision does.

Computer vision in marketing images

Computer vision is where the current golden age of AI research begins. It powers a class of algorithm called neural networks with graphics cards, which are normally used for playing computer games. Computer vision helped computers to move from unreliable Optical Character Recognition systems, to correctly identifying dog breeds within the space of 5 years.

Seemingly overnight, Google updated its photo album app for Android phones to identify what was in the image. From the chaos of a flat list of tens of thousands of images, your phone became a searchable image library.

Computer vision systems now help doctors diagnose diabetic retinopathy, which currently causes thousands of cases of preventable blindness across the world, with an accuracy better than expert doctors and within a fraction of a second.

These two examples are important and real-world use-cases for AI, but they focus on very objective tasks. With the help of computer vision we are now able to answer if this dog is a chihuahua or if certain MRI needs to be referred to a cardiologist. However, how about getting answers to subjective things?

Imagine being able to tell whether an image would make someone happy. Are computers able to provide such answers?

Will marketers be able to know what images appeal to their audiences with the help of computer vision?

Can a machine see beauty?

Marketing images can machines see beauty

The combination of AI and social sciences like psychology is a hot area to watch for marketers, especially in image analysis. It allows to predict image attributes that are appealing to audiences, segments and even individuals taking into consideration a more human approach.

Instead of just recognising dog breeds that neural networks have been trained to predict, there is now much more than that and marketers are the first to enjoy the positive implications of this in their day-to-day jobs.

Aesthetic quality of marketing images

While beauty may be in the eye of the beholder, that hasn’t stopped artists, photographers and scientists trying to quantify what exactly it is that makes an image beautiful. Whether it’s the rule of thirds, the golden ratio, or a powerful neural network trained on millions of human ratings, machines have steadily improved at their ability to predict how beautiful we’ll find any given photo.

Marketing images aesthetic quality

Memorability of marketing images

Scientists at MIT had survey participants play an image memory test and recorded their results to create a large dataset. Since then, scientists have trained models which predict image memorability to nearly the same level as humans.

Emotion of marketing images

Conveying the right emotion through images is a powerful way for marketers to encourage desired action. With enough data, AI systems can learn to predict whether an image is more likely to inspire certain emotions, whether it is awe or excitement, or something else.

New technologies are emerging to help marketers pick the most engaging, viral, on-brand images to power their campaigns. By picking the most aesthetic, memorable and emotion triggering images marketers can ensure increased marketing performance ahead of running campaigns.

Using AI to inform marketing image choices also saves the precious resources of time and user attention, allowing the marketer to test 100s of variants ahead of running campaigns without having to run costly A/B tests.

Can AI predict engagement? Let’s run a test!

At DataSine, we’ve collected millions of image posts and engagement data from various social media sources to power the AI of our Pomegranate platform. The platform assesses marketing images ahead of campaigns to anticipate assets which will perform the best.

Test preparations

Let’s take 6 marketing images, for example:

This poses a good test for you as a marketer: Which image would you expect to perform the best out of this group (whether it’s for a newsletter, social post or a landing page element)?

Running the test

All six photographs were run through Pomegranate AI. Pomegranate first breaks marketing images down into tags and assesses each tag’s individual appeal, which can be either positive or negative. Below is the list of tags that AI has identified as the most significant from all 6 images. Tags with a positive impact are shown in teal to the right of the graph, and those with a negative impact are shown in orange, to the left.

Table with image features contributing to image appeal

Images of warmer seasons and with vacation elements are considered to be more appealing by AI. On the contrary, more concrete features, like wood and structure, seem to demonstrate less aesthetic appeal. Interestingly, typical concert attributes, such as stages and crowds, produce a negative effect on people in general.

Pomegranate image ranking

Based on the features outlined above, Pomegranate rated the six images. The Flower image and the Pool image, both containing the themes of warmth and vacation, were expected to perform the best, whereas the Concert image was ranked as the least appealing among the images. 

Test outcome

Do these results make sense when applied to real world marketing campaigns?

A Facebook paid advertising campaign was run to test these 6 images in action. We received the following results:

Marketing images ai pomegranate facebook test

The main outcome? Images with the themes of relaxation, vacation, nature and summer increase results of your marketing campaigns! Next time you work on your campaign, try including similar styles and enjoy improved campaign’s performance. If ever in doubt, you can always check your candidate images with Pomegranate Free, a predictive image analytics tool, which allows you to assess your images ahead of marketing campaigns.

The future of marketing images is AI

People have completely different tastes when it comes to visual style. The more objective our image choices are, therefore, the better is the performance of marketing campaigns.

Using AI in selecting the right visuals empowers marketers remove guesswork and ensure that certain content will be appreciated by their target audience.

The next logical step here would be to start segmenting based on visual preferences. These first steps in understanding appeal of marketing images with the help of machine learning serves as a force amplifier for marketers to move towards more personalised campaigns.

2019-02-20T10:22:39+00:00