What it does
The KPI assesses how well an image matches the intended message of the asset. It assesses how closely the visual content of an image correlates with the key messages it is meant to convey.
Example: A chips brand is launching a new campaign to promote its Hot & Spicy flavor. The agency develops three design options aimed at conveying the product's bold taste and exciting vibe. Via Intended Message Fit analysis it was revealed that of the three designs, one version (V1) outperformed the others in communicating the intended messages.
Why it matters
Ensuring that images in marketing assets align with the intended message is crucial for effective communication. Proper alignment reinforces the message, enhances clarity, and prevents misinterpretation, directly influencing consumer behavior and brand perception.
How it works
The deep learning model analyzes the image, stripped of any text, to determine its semantic fit with the intended message. The model scores this fit on a scale from 0 (no fit) to 100 (very high fit), reflecting how well the visual elements of the image support and convey the intended message.
See a more detailed How it works explanation for video apps below.
How to improve
-
Select visual elements that have a strong, intuitive association with the intended message.
-
Consider symbols, imagery, and cues that evoke the desired association, like using refreshing visual elements for a message intended to convey rejuvenation.
-
Be mindful of cultural and contextual nuances in imagery to ensure the correct message is conveyed.
AI models used
The deep learning model is trained on more than 400 million pairs of images and corresponding consumer comments. During training, the model is tasked with selecting the correct text given an image, or vice versa. It learns to understand the underlying meaning and association between images and text, enabling it to assess the congruence between an image and short message descriptors like “refreshing” or “inspirational.”
How it works - video apps
Example result:
Explanation:
Lets say a video consists of 5 scenes and 3 labels are inserted in the briefing.
- Vertically you see the score each label gets for each scene resolving in the orange (label) score. I.e. in 20% of the scenes, label 1 was communicated above threshold (threshold = 50). So the score of 20 would be displayed in the table for this label.
- The overall score (Overall video calculation) is then calculated as shown in column E.
- Is at least one of the messages being conveyed above threshold. Here it doesnt matter if only one or all messages are above threshold.
- So Scene 1,2 4 and 5 have at lesat one message above threshold. Scene 3 doesnt. Thus, 4 of 5 scenes communicate at least one message above threshold