Best practices for using Predict

This article will provide you with best practices to get the best results using Predict.

  1. A/B testing, split testing, and comparison testing
  2. Test in context for quality insights
  3. Avoid long, scrolling screenshots
  4. Be mindful of white space
  5. Use resolutions that match the viewing experience


A/B testing, split testing, and comparison testing

If you test two assets side-by-side in one image, you see performance, like how attention falls, when they are side-by-side. If you want to test for a design that for example draws more attention to a specific area, like a brand logo, we recommend uploading separate images for each design and using the comparison and Area of Interest (AOI) features.


๐Ÿ“ Make sure to be consistent with image/video resolutions when comparing multiple images.


Test in context for quality insights

If possible, we recommend testing content both by itself and in the context it is going to be viewed. For example, if you test a banner ad inserted on a website, you will see how much attention that banner ad does/does not get. When you test the banner ad by itself, you get a more in-depth analysis of what stands out once the viewer has spotted it.


๐Ÿ“ When split testing your assets in context, keep in mind that design changes may have less impact on cognitive and sentiment scores.

For example, if you are testing a banner ad in context on a very noisy website, you typically canโ€™t change anything about the website, which limits your ability to impact Cognitive Demand and Focus. Instead, you can see which of your designs draws the most attention to your banner ad on the website by using the AOI tool.


Avoid long, scrolling screenshots

Predict analyses the whole image or frame, assuming that the viewer sees it all at once. But this is rarely the case with websites, app interfaces, or emails. Instead of uploading long screenshots, we recommend taking screenshots of each section to get true-to-life results.


Be mindful of white space

White space impacts the scores too. If you test an image of a product with a white background, that white background will also affect the score. It will lower the Cognitive Demand as white space is not very demanding. Focus will typically go up as almost all the attention will land on the product in the center, unlike in more lifelike situations. Therefore, try to limit the amount of white space and keep it consistent when A/B testing or comparing.


Use resolutions that match the viewing experience

Test in a resolution that best fits the viewing experience. You may have a super detailed high-resolution design, but if the viewer is experiencing the image on a phone, the finer details are never picked up by the viewer's eye. For the most accurate results, we recommend testing in the resolution your type of content is usually displayed in.