fbpx
Back to Blog

AI and the Future of Marketing: Predicting Ad Performance

Imagine being able to predict ads that will convert before you ever put them live? How much would that impact your business?

Published by Dr. Tadas Jucikas, Founder and CEO at Genus AI

For the first time ever in 2021 social media ad spend has exceeded that of paid search [1]. Online video is the fastest growing ad format propelled by performance focus from e-commerce [2]. However, for e-commerce D2C brands growth via the social channels is not as easy to master as it seems.

D2C brands need to create winning image and video ads and turn to creative agencies or videographers to achieve that. It can cost tens of thousands and take weeks to produce a single video ad. In the ever changing world of Facebook and Instagram ads you always need new ads and such cost and time commitment are hard to come by.

Similarly, in order to build efficient audiences the brands need data science resources to evaluate their customer data and build lookalike audiences from their own customer base rather than more ephemeral pixels which have been heavily disrupted by the changes introduced in iOS14.

And lastly, running efficient social media campaigns requires experience and channel understanding usually found with expensive agencies which sometimes don’t even speak with smaller brands prompting them to build capabilities in-house.

However, Artificial Intelligence has the potential to make every brand, no matter their size or budget, an absolute powerhouse in terms of ad content creation, audience discovery and campaign management.

Today, Genus AI Customer Growth Platform offers such AI based capabilities to savvy growth marketing teams who value control and efficiency [3].

End-to-end system of AI driven video generation, content scoring and audience creation
Try out Genus AI platform for free at: https://genus.ai/.

It is not often that a feature release causes a huge ripple effect across the entire platform and has a potential to change how entire teams work. However, this month such a feature was released by the Genus AI applied AI team.

It is now possible to predict CTR* (click-through-rate) for image based ads instantaneously.

This predictive model was trained on the historic performance of tens of thousands of ads which received over $62m in ad spend. Accessing this knowledge allows brands of any size to remove the guesswork from the ad creation process. Now, any new image uploaded to the platform can be instantaneously scored with a CTR model opening up never before seen ad sourcing capabilities to brands (e.g. score an entire library of tagged posts, score thousands of stock images and score whole photo shoots to source ads).

Image CTR model can instantaneously evaluate your ads and rank them in terms of their likelihood to generate click-throughs.
Image CTR model can instantaneously evaluate your ads and rank them in terms of their likelihood to generate click-throughs.

Imagine a world where you know if your ad is going to be clicked on before you ever run it and spend any of your budget.

How much of an impact would that create for your business? Some brands who used the model to score large image libraries saw a tremendous impact on their CPA and ROAS, decreasing CPA by 30% and doubling their ROAS.

As the team worked through the model, it was tested on thousands of ads (4.2k to be specific) where the actual CTR was known and was recorded. The model has never seen these ads before and scored them returning a number between a 0 and 100. Here are the results when calculating the average actual CTR by every 10 points from least likely to most likely to have high CTR.

CTR score range
Benchmarking CTR model results on randomly selected 4.2k ads that the model has not seen before.

If you want to reach higher ad CTR all that you have to do is to make sure your ads score at least 50 on the platform. If you really want to drive CTR up, aim for at least 80. This analysis can be done on any number of images at any time — there is no limit to how many images you can evaluate.

What we noticed when evaluating the test ad library is that CTR would sometimes score high for ads which don’t look that great. For example, they would have a uniform background and text offering a discount or some other announcement. While such images are indeed likely to generate clicks they are not ideal to be used as ads.

Luckily, the Genus AI platform already has a score called Ad Appeal which scores how likely an image is to be liked and engaged by people. This score allows us to identify which image is appealing and will thumb-stop and generate an engagement.

Ad Appeal score
Benchmarking Ad Appeal model results on randomly selected ads that the model has not seen before.

Ads scoring over 70 are more likely to generate a larger number of likes and engagement.

Such critical metrics as engagement or CTR are all very important but are just precursors to what really matters for most brands: how much it costs to generate a purchase.

Cost-Per-Acquisition or CPA is the final metric that a lot of brands track very closely. It is interesting to observe the resulting average Cost-Per-Acquisition (CPA) for different Ad Appeal and CTR ranges.

Average CPA for different Ad Appeal and CTR ranges for 4.2k ads. The highest average CPA (Low AA+Medium CTR) is compared to the other ranges.

The highest average CPA is observed for images that score Low in Ad Appeal and Medium in CTR. These lower quality ads deliver inefficiencies and should be avoided when creating campaigns.

The most efficient ads are Medium in Ad Appeal and High in CTR as well as High in Ad Appeal and High in CTR. The difference is staggering — your CPA could drop to ~40% of what you would get in the Low Ad Appeal + Medium CTR scenario. In other words, if your ads happen to be in Low AA and Medium CTR category and your CPA is $100 — it can be as low as $41 if the ads are created with High CTR and at least Medium Ad Appeal score.

If you want to decrease your CPA — select ads that are High in the CTR score and at least Medium in the Ad Appeal score.

Here are some examples of how efficient ads with High Ad Appeal and High CTR look like:

Examples of ads that have High Ad Appeal and High CTR — most likely to have the lowest CPA.

They have higher contrast, engage the viewer and seem to ‘jump out of the screen’. Similarly, here are some ad examples with Low Ad Appeal and Low CTR:

Ads that have Low Ad Appeal and Low CTR — most likely to have the highest CPA.

As you can see visually, they are quite different: highly curated product shots versus images resembling stock footage.

To an untrained eye both sets of ads can look perfectly acceptable. However, with the help of the Genus AI platform brands no longer need to create in uncertainty in terms of how the ads will perform.

Brands now can create their growth campaigns and high performing ads with confidence.

It is clear that there’s a tremendous amount of spend in digital advertising and that certainly is not going away. There are clear shifts going on in the industry as well: overhaul of the agency model, businesses are looking for accessible tools to produce, edit, source content, and searching for proven ROI, continued momentum in the UGC space and in influencer marketing.

Artificial Intelligence is transforming marketing and how companies grow.

At Genus AI we are committed to helping brands create ads with confidence and accelerate their growth via social channels. The platform that we have built directly capitalizes on several of these trends by leveraging the latest AI innovation to help brands of any size grow and create with confidence.

[1] https://www.mediapost.com/publications/article/365393/zenith-online-video-to-lead-2021-global-ad-spend.html

[2] https://www.mckinsey.com/business-functions/strategy-and-corporate-finance/our-insights/five-fifty-the-quickening

[3] https://genus.ai/

* CTR is defined as ‘Total Ad Clicks’/ ‘Times the ad was “seen”’

Share