Design in the Age of Data and AI

There is no shortage of data. User behavior, research and A/B testing are easily at hand — or could be easily at hand with a little effort. Additionally, AI and machine learning are a reality thanks to the abundance of data. The benefits to marketers are huge, as segmentation and delivery can be more precise than ever. But does all this data benefit the design process? Does data-driven design force us to look backward, rather than move forward and create something totally new?

I can’t answer that question, but if you look at the history of cell phones, the answer is: don’t rely on the data. In the early 2000s, research showed that people didn’t want and wouldn’t buy a touchscreen phone. The data was wrong.

We see benefits every day of using data to target, segment and deliver great creative. Robust analytics help us review and optimize results. Automated optimization helps us see anywhere from a 2 to 2% lift in email marketing, display advertising and search marketing. But what about design? Can data be used to create something new, rather than merely improve what already exists?

CMD recently conducted an experiment on Google’s ad platform in which we let Google’s AI system create and manage ads. Here’s how it works: Essentially, you provide a URL and Google mines the destination for images and copy to create its own ads. Google has had real-time campaign optimization for years, and we occasionally use automated Google Ad Words with … okay results. This time, though, we wrote some copy and created images and ran these ads side by side with Google auto-created ads.

The media buy was only $10K, and the method wasn’t very scientific, but what we found was that the automated ads performed much better on mobile than the human-created ads. On desktop, however, results were pretty even, with no real difference between human-created ads and machine-created ads.

Do I believe that AI can build better creative than a human? No. This was a simple test of some display ads that still required a human to create the base content. But I do believe that AI for marketing is progressing. Automated systems can crunch the data quickly to tailor content and offers on the fly without a human working through complex business rules and manually creating the many permutations. I also believe that data can guide designers, but the story has to come from imagination. The question is, will that be human imagination — or a machine’s?