AI-Powered Concept Optimization [Beta]
What is AI-Powered Concept Optimization
The Activate It AI generated concept revision feature allows you to see what your concept description might look like if the feedback from your survey was applied directly onto the concept by an AI trained by Zappi.
The revised concept will also include an explanation from the AI of why it changed the concept in this way and visual recommendations outlining what elements to adjust on your packaging. You can generate AI revised concepts for the general category audience as well as for a specific target group (profile).
This is a new idea we are experimenting with and is available as a free feature in Beta.
Where to find Concept Optimization?
Activate It AI generated concept revisions are currently only available for Activate it 2.0 FMCG & QSR concepts as part of the free open Beta. We are experimenting with making it available across a broader geographic portfolio but this is dependent on LLM's evolving capability to handle other languages.
You can download the AI concept revisions either by clicking on the banner in the top part of your concept analysis / reporting output
or download it straight from your concept tile (click on the three dots in bottom right of the tile and then ‘Generate revised variations’)
You can choose the target group for which you want to generate the optimized concept. If you wish to generate an AI optimization based on feedback from a specific target group, select a profile.
What can I do with it?
The purpose of the AI generated revisions is to give you a head start on concept optimization. We encourage you to use it as a great starting point for revisions that helps you tweak the concept more effectively based on your judgment.
We’ve seen significant improvements in performance of concepts optimized by our AI agent using Zappi data. When testing the AI optimized concepts with consumers, 20% of key concept metrics saw significant improvement vs original concepts.
We know combining AI-driven concept optimization with human insight leads to even better results. Here's why:
- AI's Analytical Power: AI swiftly analyzes data to identify optimization opportunities.
- Efficiency: AI accelerates the optimization process, saving time and resources.
- Data-Driven Insights: AI uncovers patterns and trends for informed decision-making.
- Human Expertise: Humans provide contextual understanding and industry knowledge.
- Refinement and Adaptability: Humans apply creative thinking to refine concepts for maximum impact and adjust strategies based on changing needs.
- Quality Assurance: Human oversight ensures optimized concepts meet desired criteria.
So experiment with it! Combine AI's analytical prowess with your expertise, and send us your feedback and suggestions so we can further improve the model. Try it more than once and you'll see how it can create something new each time.
How does it work?
In order to revise the concept, we identify the dominant themes in consumer open ends (Likes & Dislikes) and analyze performance on selected KPIs - Clarity, Breakthrough and Brand Believability as these are the three areas where improvements are not easily identified from open ends. Then a LLM (Large Language Model) applies the results of this analysis to the revised concept, addressing deficiencies and amplifying the elements that were appreciated by consumers. These revisions are applied to the text description only (we anticipate making image adjustments in future versions).
We then instruct the LLM to evaluate the revision from the perspective of the target audience - the ‘synthetic respondent’.
What is the synthetic respondent?
For the synthetic respondent feedback, we are assigning the LLM a persona in line with the concepts target audience, and asking it for its impressions of the revision. In future we will have a full survey of synthetic respondents and we are also playing with hybrid models of synthetic and real respondents but again, this is our first iteration, tell us what you think!
Why is it in Beta?
Our new feature is a proof of concept for an AI concept optimization agent. In the future you will be able to push it in different directions but we have to start somewhere and this is our first feature. There are many more to come!
Large language models are groundbreaking, and constantly evolving. Their occasional inaccuracies, or 'hallucinations' are part of the learning curve—not just for the AI but for us as we integrate your feedback to enhance the feature's performance.
By participating in this beta, you're helping shape a smarter, more intuitive system. We're committed to transparency and improvement, working towards a future where our technology serves you with even greater precision and personalization. Thank you for joining us on this journey of innovation and growth!
What about security?
Your privacy and security are our top priorities. We adhere strictly to Microsoft Azure's comprehensive data security policy to ensure your information is safeguarded at all times. This means that any data you provide while using the feature is encrypted, securely stored, and never shared without your explicit consent. Your data (inputs), and your AI revised concepts (outputs) are NOT available neither to OpenAI nor to any third party.
Who owns the generated content?
When you use our feature to generate content, you have full ownership of the generated concepts. We're here to inspire and enable your creativity, not to claim it. Your success is our success!