January 22, 2024
For creative teams to get what they need from AI, they need to collaborate much more closely with their analytics teams.
By Alex Collmer
Generative AI has become a mainstay in our daily lives. Whether using ChatGPT to create a workout plan or Midjourney to generate an image, there is one key learning that becomes apparent quickly: the input, or the prompt, is what matters most to a successful outcome.
The importance of inputs in getting the most out of technologies is not new.
Take maps, for example. My wife and I both have Google Maps on our phones. If I type in "Vermont" while she enters "57 Schuss Pass, Waitsfield, VT 05673," she is going to get to Mad River Glen and I will not.
Maps technology is a commodity. We all have access to it. It is what we put into it that delivers a positive or negative result. The same will be true for generative AI.
We all have access to ChatGPT, Midjourney, Firefly and the many other options for generative text, images and video. Like maps, the technologies are a commodity – It is what we put into them that delivers different results.
For advertisers, the question of putting in the best possible prompts that will lead to more performant content will be absolutely critical.
Marketers need thousands and thousands of creative assets in all different shapes and sizes for different platforms and audiences, but simply filling the slots will not be successful. They need to perform and they will need to improve over time.
AI can help with this, but only with inputs that are driven by continuously updated performance information.
While data lives in many places – in campaign reports, on advertising platforms and even in the creative assets themselves – it is often siloed. The analytics teams typically own this data , but historically have not had close working relationships with creative teams.
As a result, the data is often underutilized or unused.
For creative teams to get what they need from AI, they need to collaborate much more closely with their analytics teams.
Alex Collmer
A prompt without data
There are two places where AI is making its way into the creative process.
The first iteration of a creative idea usually comes from collaboration with a brand, creative talent, late nights, mood boards and inspiration.
The role of generative AI at this stage is to trigger new ideas, provide insights on fertile grounds to explore, do rapid prototyping and add variety.
After the initial concept is born, creative teams spend a lot of time trying to figure out if their creative concept works, fitting it for different audiences and placements.
Too often, creative concepting, iteration and changes are made with little data. There are sometimes reports from campaign tests or focus groups, but rarely do those reports provide the tactical information that can help creative teams confidently make changes at any level of detail.
The same brilliant idea and message can perform wildly differently on each platform with subtle tactical tweaks to its execution.
The role of AI here is to provide these tactical insights, automate their execution, scale the concept and iterate quickly.
There are a growing number of examples to turn to for anyone looking to assess the potential impact of this approach, and to see how to get started.
Kenvue is one. After nearly a year of not advertising on Meta, Kenvue used AI to help.
First, the marketer scored its creatives to improve adherence to brand requirements and Meta platform best practices. Then, it used AI to analyze key performance drivers in its Benadryl creatives to inform their approach.
The second place where AI is already helping is in scoring content as it is being produced.
Many marketers have found that as they scale the number of assets needed, they have lost a degree of consistency and quality.
With hundreds of agencies and production companies doing work, many assets are created that do not fit brand guidelines, DE&I goals or even basic platform and audience best practices.
AI can help generate scores for each and every asset, so marketers stop putting the wrong content in the wrong places, and wasting ad budgets as a result.
Both of these processes benefit from creative data. All of the insights that analytics teams have can help creative teams get more from AI, which in turn helps creative assets perform better.
Analytics tools can compare sophisticated creative data to a wide range of audience behavior KPIs such as media metrics including CPMs, engagement and view-through, brand KPIs comprising brand lift, recall and favorability, and even offline sales and other measurement tools encompassing MMM models.
In doing so, they can generate insights which can be translated into data-driven prompts, leading to more performant AI outcomes.
Companies that prioritize an AI-based creative performance infrastructure will get much stronger, more actionable signals.
Knowing what details within a creative drive performance – such as how the product is presented, presence of talent, music, text, emotion and color – can help teams understand more about the audience, where to place the ad and how to change the creative to resonate more.
Learning from each other
To connect the dots and increase ROI next year, analytics and creative teams will need to work more closely together.
Creative accounts for the majority of a campaign’s performance (60 percent). Integrating AI-based creative data throughout the content supply chain increases marketing efficiency and performance.
Collaboration does not need to be complicated, but teams do need to commit to the process.
Ideally, a central stakeholder is designated to facilitate more interaction and sharing between teams.
A good start is to create a small “center of excellence” that includes leaders from the brand, analytics, creative and media teams to discuss how AI could inform different processes in the form of insights, automation and scale.
Discussion and exploration will uncover many opportunities.
For example, analytics teams have AI-supported insights that creative teams are not getting access to, while creative teams might have design processes that are heavily manual that could be automated with AI.
Ideally, the team shares a single performance goal to encourage testing and process improvements.
IT IS CLEAR that AI can be valuable individually for both creative and analytics.
What we now need to embrace is how AI can also bring creative and analytics closer together.
Alex Collmer is CEO of VidMob, New York.
Share your thoughts. Click here
