Frankenstein Ads: Use AI to Find the Right Copy and the Right Image for the Right Audience

Optimizing your conversion funnel is about more than just your website design. The typical digital journey can cover a variety of other touch points, including social media messaging, live chat conversations, and the advertisements that attracted users to your site in the first place.

Conversion rate optimization (CRO) may still seem a relatively new discipline to some, or a buzzword shoehorned into job descriptions at companies keen to maximize the return on their digital assets.

But it is about more than just running A/B tests on your page layouts. In fact, CRO should be seen more as a mindset; a series of principles that applies to your entire sales journey. The core principles of testing, measuring, learning and improving can generate positive improvements across the board.

Testing your website designs is, of course, an essential ingredient in a successful optimization programme. It is at the heart of Sentient Ascend’s core product, the ability to test out an unlimited amount of design combinations and iterations to help improve your conversion rates.

This testing mindset can be applied to all of your online activity to help improve the entire digital experience, optimizing every stage of the sales funnel. Today, we’re going to look at the benefits of testing your advertisements, to help take the guesswork out of your creative campaigns.

The anatomy of an online advertisement

Testing is already quite a common practice among PPC execs. Or at least it should be. It might consist of trying different headlines in your search campaigns, using a variety of images in your Facebook Ads, or testing font sizes on your display creative. But think about how many elements go into a typical online ad and the list of possible changes that can be tested quickly grows.

For search ads, you’ve got the headline, body copy, use of capital letters, ad extensions, sitelinks, display URL, final URL and landing page. Your display ads can vary by format, with a single image, slideshow, video or GIF. The image or content can be changed, as can the filter or color scheme, the copy, the text size, text colors, and the size and shape of the ad. Then for Facebook, there are even more factors to consider, such as the variety of formats now at your disposal.

The effectiveness of all these ads is also impacted heavily by your audience targeting, whether that’s your keywords and bid adjustments, retargeting settings, or your targeting and placements for social and display campaigns.

Finding the killer combinations

When you consider all those variables that go into a typical campaign, various factors can influence its success. Traditionally, an A/B testing approach to optimizing an ad would only change one element at a time, otherwise, how would you know which change had the biggest impact?

So for a Facebook ad, for example, you might want to test out two different images to start with, running both variations at the same time to see which drives the best results. After running the test for a couple of weeks, you choose a winner. Then you want to test something else, so you try out a different headline alongside the winning image.

But what if the new headline you test actually would have worked better with that original image that you’ve already discarded? Unless you take a multivariate testing approach and try the different combinations, then you’ll never know, and you could be missing out on your golden goose. This can lead to a seemingly endless number of potential combinations.

Say you want to test five different headlines, with five different images, plus five different examples of body copy, that gives you a total of 125 different ad variations to try out (5 x 5 x 5 = 125). And that’s before you’ve even started to test different audiences.

As anyone who has ever set up a single Facebook ad will know, trying to build 125 of them will take a long time, while it will take even longer to run them all and reach any statistically significant conclusions about which is the ‘best’ variation. But figuring out how different variables and elements work together can be key to driving growth, and can often throw up some surprising results. It’s also very likely to be the case that there is no single ‘best’ ad, as what works for some people, might not work for others, Different audience segments or users can respond differently to different elements.

Responding to external influences

Advertising does not exist in a vacuum. As well as the variables that we have looked at for what goes into the ad itself, there are also various ‘outside influences’ that can affect the impact of your campaigns.

Think about the weather, for example. It can have a huge impact on a person’s mood, mindset, and priorities, even affecting their potential to respond to messages or buy certain products. Travel companies often see higher conversion rates when it’s raining. There are email marketing tools that will trigger a particular campaign geo-targeted to people in areas when it starts raining, or using dynamic content blocks that change depending what the weather is like at the exact time a person opens the email.

It’s easy to see how the same principle could apply to your ad campaigns. If you’re running display ads then the creatIve could change depending on the weather or time of day a person sees it.

All these external factors should also be considered when analyzing the results of your ad tests, in case they might have influenced the results. That image you discounted from your social ad campaign during the first round of testing might actually perform better when the weather turns nasty.

The opportunity for AI to shape the future of advertising

With all these factors to consider and elements to test, how are marketers embracing the latest technology to make their campaigns as effective and efficient as possible? Sentient Ascend has pioneered the use of evolutionary algorithms and artificial intelligence to revolutionize how people approach on-site optimization testing. And there are plenty of use cases for how similar use of machine learning can improve the accuracy, speed, and success of digital advertising campaigns. AI has the capability to help with everything from redesign ads at scale and writing optimal ad copy, to hyper targeting and personalizing the content to drive better engagement.

Audience targeting is a crucial ingredient for effective campaigns. And it is an area where AI-powered platforms can help to quickly match audiences to the most effective ad content at a speed and scale that was not possible with traditional demographic targeting. It is a technology that advertising giants are already investing heavily in.

Watson, IBM’s artificial intelligence service, has already begun delving into many of these areas. Its ad platform has experimented with content creation, audience targeting, and real-time optimization. Google’s AdWords platform also has machine learning deeply embedded into its offering, with the search giant steering all its customers towards the algorithmic optimization services for everything from ad copy testing through to bid strategies based on using AI to identify people most likely to purchase, bidding more for those warmer prospects in the ad auction.

A host of other services are also appearing in this space too, helping to take the grunt work out of testing and optimization. For example, Zalster helps split test the objective of your ads, taking a full-funnel approach to identifying prospects at every stage of the buyer journey. Trapica promises to optimize your bidding process, while ReFUEL4 is another service that uses AI to help with the heavy lifting of refreshing creative and copy, tackling problems such as ad fatigue.

AI platforms are also attempting to ease the complexity of running cross-channel campaigns, adapting the content and creative to suit the medium, as well as the audience and messaging.
Albert is an AI tool that combines machine learning with predictive analytics and natural language processing to create an autonomous advertising solution, servicing creation, execution, testing, and optimization.

Facebook itself is now rolling out a ‘dynamic creative’ feature which delivers ‘the best combinations of your creative assets’, running different combinations of these assets across audiences. It is not completely hands-off, as you still need to input the building blocks for the creative, such as the titles, images, and videos, text options, descriptions and CTAs.

That human involvement is a crucial point when it comes to embracing the potential of AI. Despite the Hollywood-style fears of robots taking everybody’s jobs, effective advertising and optimization still need a creative and strategic input from human minds. But what the advancement of AI-powered platforms and tools will allow for is testing, targeting and learning at a speed and scale which was not possible before.

Finding the perfect combination of ad content and audience targeting may have always seemed like a pipe dream for advertisers, but there is now more data and more accurate insights available than ever before, to make that pipe dream more of a reality.