It's taken us a long time to figure out how to machine-generate meaningful, new ad ideas from the simplest inputs in just minutes. And then use them to help brands 2x ROAS.
Here's what the ads look like, and what we learned:
1. It’s not about variations. It’s about new ideas
Testing small variations of the same ad will have a small impact. Testing large variations, or better yet trying totally different ideas, will have a large impact. If you test enough, both approaches are guaranteed to deliver. But you have a finite testing budget and hours in the day. What do you do?
With machine generated creative, you do both.
Variations leverage statistical anomalies in what resonates with people, yielding small changes in outcome that might be magnified over time or as the number of people scales. Google famously tested 41 different colors for search results links, and the shade of blue they settled on earned them an extra $200M a year in ad revenue. But most brands are not Google, and rigorously testing every micro factor in your ads will burn a lot of ad budget and may only deliver a few percentage points of uplift at your scale.
This is why, for most brands, testing whole new ad ideas is so powerful. If each new ad idea is distinct enough to resonate with a whole new group of people, Facebook can find others like them easily and scale your spend and ROAS. Job done. And while this is happening, MGC is tracking which micro factors are working consistently across your ads and upweighting them in every new ad you generate.
2. Ads are just noise unless they are meaningful
In the pursuit of new ideas, creative variety and ROAS it’s crucial to remember that you are still a business trying to find compelling ways to build relationships with your customers. It’s very easy to variate your ads too much and find yourself sounding like a competitor or worse, miss-selling yourself or undermining your brand.
With machine generated creative, it’s now possible to generate not just mindless variation but to genuinely be creative, with unique ideas, at scale.
Your brand guidelines and product facts are set up and locked. Meaningful creation then starts with your copy. You can automatically generate new ways of describing your products and pitching your purpose. Once you have these you can automatically generate video ads that have real narrative - a story that grabs attention, lands a benefit and drives towards conversion - all by slicing and remixing your existing assets in alignment with your story.
3. Briefs have to be simple while offering control
Creativity is a messy and wonderful business. You love creating something new but not sorting through hours of footage to find that perfect scene. Somebody else has the font files on their laptop, and they’re on holiday. Nobody has the working file from last summer’s campaign at all.
Machine generated creative works for real people because it starts from the basics. Just key in your website URL and your carefully crafted branding, logos, fonts, colors and tone of voice is filled. Upload just product imagery and video and all your assets are automatically tagged. At this point you have everything you need to make ads in minutes.
Perhaps you want to just try some new ideas. Or you have a specific proposition you want to test on your bottom-of-funnel LALs. Controlling generation is just a matter of giving some copy examples of what you’re trying to say and clicking a button. By choosing the ads you like from what's been generated, you close the loop on control.
4. Minutes are the speed of collaboration
Getting to any strong creative output, including your weekly batch of brilliant ads, is always a collaborative process. Creativity is subjective. You won’t like everything a machine generates just like you don’t like everything your agency presents to you. You're a lucky human if you like half of your own ideas!
One of the reasons machine generated creative works so well is that it isn’t totally machine generated. You provide the brief. You make the edits. It learns from you, what you like and what gets results for you. One key to making the MGC process collaborative is fast generation - just seconds for copy and minutes for video. While you wait, you’re collecting your thoughts but not tapping your fingers. Another key is making the back-and-forth effortless. There aren’t dials to turn or knobs to twist, you just provide your assets and type in the copy ideas that guide the ads you want to make. You learn from the AI and it learns from you.
5. Nothing matters if the ads don’t work
At the end of the day, the most frustrating part of making ads (for both the performance marketers and the creatives!) is ads that don’t work. The thing is, because testing is the only guaranteed way to consistently improve, it’s a law of nature that most ads made will fail. Cue a lot of wasted time, wasted resources and spent emotional energy.
Machine generated creative gives you a secret weapon. Not only does it make creation fast and low cost, but it shortcuts the results by predicting what might work based on what’s worked in the past. This has 3 big advantages.
First, you are no longer spending all your time and money making ads for most of them to fail and be paused, never to be seen again. You generate an ad in 1 minute, run it, and if it fails it's no huge loss. This is an energising feeling.
Second, you save money on your testing (and your customers' eyeballs) because you don’t need to test every variation. Pencil remembers what worked last week and last year and guides you towards new white space to test and away from things you’ve already shown don’t work.
Third, every single ad you run just helps you hone in on what really matters for your customers. After all, isn’t that what advertising’s all about?