Man against the Machine — Creatives Make or Break you

Promoting mobile apps with display ads is hard. Users mostly ignore your ads, conversion rates are naturally low and you are competing for people’s attention with other apps (and ads). Showing the right ad creative is one of the very few variables that marketers have full control of, yet the importance of a sensible creative strategy is often overlooked. Designing and testing as many creative variations as possible should be the norm, but the process is often limited to just a few options per set (and sometimes there’s a lot of sets — think different countries/products/special dates) as design resources are usually very limited.

When we analyzed the Click-Through Rate (CTR) evolution for a creative that has been used for a couple of weeks without getting changed, we found that the CTR almost halved after 5 weeks.

But when we introduced new creatives…

The impact of the refreshed ads is significant and that is the tip of the iceberg in terms of what you can do with creatives.

For starters… an ad is not a single image 👀

An ad is made up of several creative elements:

  • Logo
  • Design
  • Colors
  • Images
  • Copy
  • Font Style
  • Font Size
  • Call to Action Button Color

And by using dynamic creatives you can treat each of these elements as an optimization variable.

So, are some more important than others?

We conducted a series of A/B test experiments to find out:

  • If the change of an individual component of the creative theme had an impact on the user behavior
  • If certain elements had a bigger impact
  • If the difference between standard and customized elements had a significant impact

Each experiment changed only one particular component of the creative theme.

Experiment 1: Brand Colors

We run two sets simultaneously. The two designs were identical (same image, text and text styles) except for the color palette.

The main color of the palette (the brand’s most representative color) was used as the main color on the Treatment design.

Image for illustration purpose only. The tests were conducted using real brands.
Table 1: Brand Color Experiment Results. Source: Internal Jampp Data

The results showed that the ads where the brand colors were more prominent had 17.18% better CTR than the control ads. This suggests brand recognition impacts conversions. Though this may also have to do with this brand’s particular palette. If your app’s main color is black, you may find alternative colors work better.

Experiment 2: Text Length

Image for illustration purpose only.

Modifying the copy of the ad whilst keeping all other elements equal seems to have no significant impact on performance. The ads with less copy did performed slightly better though (3.77% higher CTR)

Table 2: Text Length Experiment Results. Source: Internal Jampp Data

Experiment 3: Images

Image for illustration purpose only.
Table 3: Image Experiment Results. Source: Internal Jampp Data

While CTR was not greatly affected, conversion rate did vary per image. In fact, using different images seems to be what impacts the most in terms of user behavior (more than changing colors, copy, copy length or text styles). We saw up to 5x higher CVR in some cases. The difference is not always that significant but images definitely do have an impact

Experiment 4: Design by Vertical

After we tested the individual ad components, we decided to compare performance between themes. Did certain designs perform better? Were there any trends per vertical?

We tested multiple themes for different verticals across a wide range of advertisers. The tests were performed under a multivariate testing framework, where the only thing that changed between each variant was the theme itself.

At the end, the different designs were ranked from highest to lowest according to their respective CTRs in order to pick the ones that performed better for each particular vertical.

Some of our customizable themes 😎
Table 4: Theme Experiment Results (Themes can be seen in the previous image). Source: Internal Jampp Data

While these results are not conclusive, in the sense that the Circle theme won’t necessarily perform better for every social app out there, they do raise a few interesting points.

Follow the Data

The “prettiest” ad won’t necessarily be the best performing one. Food delivery apps often have mouth-watering photos of delicious food and yet, in the experiments we conducted the ads that performed better weren’t the ones featuring dishes more prominently.

It’s important to test different ads and see what resonates with your users.

Don’t Settle

Often times advertisers will find a format/design that performs well and “stick to it”. However, as we mentioned at the beginning of the post, all creatives (no matter how awesome) will have a CTR drop over time. It may well be that the Screenshot and Typographic ads performed better than the Circle and Screenable ads on the Food Delivery campaigns because they were more different than the ads that were running before. If you are not creating lots of different ads and updating them often, you are missing out.

Images and Colors

So, according to the experiments we ran, images and colors are the ad components that have the most significant impact on CTR. Alternatively, font style and font size barely moved the needle.

Last but not least…

Dynamic Creative Optimization (DCO) a.k.a Let the machine do the hard work

When it comes to mobile ads… creativity, design and branding play their part. For sure, but ultimately it’s a numbers game.

You want to test, and test often and use the results to create compelling messages. But testing is not for amateurs.

An ad has several creative components, when you analyze them in relation to other variables like device characteristics, publisher apps where we find those users, etc and multiply that by the number of impressions… that’s a lot of data… Enter the machine 🤖

With Jampp’s Creatives Lab, we enable the creation of multiple fully customizable ads in line with brand requirements.

Creative performance is then analyzed and optimized using machine learning. We use a refined version of Multi Armed Bandits (MAB) which allows us to select the natural winners without sacrificing the exploration process, as you would with a Greedy algorithm.

The “thou shall test creatives” is not a new “commandment”, what’s new is what we can test and how we can use those insights.

You can read more about our Creatives Lab here, or get in touch!