Neurons AI Metrics & Explanations

This article explains the cognitive & sentiment scores you can generate with Neurons AI.

•  Focus score
•  Cognitive Demand score
•  Memory score
•  Engagement score
  

Legacy AI metrics

•  Clarity score
•  Engagement score

---

Focus score

Focus is an index of focused attention in an image or a video.

In other words, it is an index of how large a portion of your asset draws attention. It also indicates the level of consensus between viewers on where they will spend their time looking at an image or video frame. If your image or video has many items that draw attention, people are less likely to focus on any single part of the asset.

How do I interpret Focus scores?

• Low Focus scores (0-24) are achieved when attention is scattered across the image or video frame. In this case, viewers look at many different areas of the asset, but will most likely not see them all. When attention is scattered, you have very little control over which areas and elements your audience will see on your asset.
• High Focus scores (75-100) are achieved when a single or very few narrow areas draw most of the visual attention. In this case, attention is very focused on one single element or just a very few elements. Your audience will very likely see this or these elements when exposed to your asset.

How is Focus calculated?

The Focus score summarises the attention distribution, as characterized by the Attention Heatmap, in a single number computed in a deterministic fashion from the attention distribution. In a nutshell, we first compute the sum of all pixels in the Attention Heatmap, and this is subsequently divided by the number of pixels in the map. We then convert the resulting number to a percentage and apply an inverse sigmoid transformation to yield a Focus score in the range of 0 to 100.

In layman's terms, the Focus score is dictated by the area occupied by the spread of the attention hotspots or "blobs". As such, the Focus score is higher when the attention blobs occupy a smaller fraction of the image. Conversely, if they are more diffuse and spread out, the Focus score will be lower.

Theoretically, a Focus score of 100 would have all of the attention focused on one pixel. In contrast, a Focus of 0 would mean that all the attention is evenly divided across the entire image or frame.

You can also think about the Focus score as the index of how much people will look at specific areas of your asset in relation to the total size/area of your asset.

 

  Cognitive Demand score

Cognitive Demand measures the amount of visual information viewers have to process while looking at your asset.

When images and videos are more visually complex, the corresponding Cognitive Demand score will be higher, resulting in increased perceptual load. Complex images might heavily draw on our working memory, making them very difficult to remember, even temporarily.

The score is based on a measure of visual complexity and entropy, and has been independently validated by correlating the score with brain-based (EEG) measures of Cognitive Load.

How do I interpret Cognitive Demand scores?

We use Cognitive Demand to understand how an image or video will be received. Ideal Cognitive Demand scores may vary for different contexts and stimuli.

• Low Cognitive Demand (0-24) indicates that an image or video frame is simple and very easy to process, however, this can make the viewer spend less time looking at your asset as it is very easy to process.

• High Cognitive Demand (75-100) makes it harder to find single elements in an image and makes it harder to understand the image as a whole, and it requires more processing power of the brain (Cognitive Load). High Cognitive Demand can be more stressful and can leave negative emotional effects and makes visual information harder to remember.

How is Cognitive Demand calculated?

For over a decade, Neurons has been collecting data on the amount of information in consumer experiences  - especially with methods such as eye-tracking and EEG brain scanning.

Since 2013, we have been working on an automated measure of visual complexity and how this relates to cognitive load (i.e., the overall workload of your brain). 

We have improved this automated measure by drawing on a well-established mathematical description of entropy, i.e. the difference between pixel intensities in the image/frame, as quantified by a dissimilarity measure. In essence, the latter provides information about the texture or pattern variations in an image.

A higher dissimilarity value suggests a greater level of variation or heterogeneity in the image's texture. This indicates the presence of different types of textures or patterns with distinct intensity variations, thereby resulting in an increased Cognitive Demand score.

On the other hand, a lower dissimilarity value implies more similarity or homogeneity in the image's texture. This implies a relatively uniform or smooth texture with less variation in intensity, yielding a decreased Cognitive Demand score.

 

 

Memory score

Memory Score is an ad recall probability metric that predicts how likely viewers are to remember an image or video asset after brief exposure. It measures the asset's ability to create a lasting impression, even after viewers have been distracted.

How do I interpret Memory scores?

Memory Scores range from 0 to 100:

• Low scores (0-40): The asset is less memorable and viewers are unlikely to recall it after exposure.

• Medium scores (41-70): The asset has moderate memorability and some viewers may recall it.

• High scores (71-100): The asset is highly memorable and most viewers are likely to recall it after exposure.

Key factors that enhance memorability include:

• Presence of people and faces

• Strong emotional impact

• Visual saliency (contrast, bold colors)

• Clear focus on key objects or products

• Prominent, meaningful text

• Simple designs with clear focal points

For videos, additional factors include:

• Vibrant colors and high contrast

• Unique and dynamic content

• Semantic richness (conceptually engaging scenes)

How is the score calculated?

The Memory Score is calculated using Neurons proprietary advanced machine learning models trained on extensive human-based studies:

• Data collection: Participants view images or video frames briefly in a rapid succession task.

• Recall measurement: Participants indicate when they recognize a repeated image, with both accuracy and response time recorded.

• Model training: A deep learning predictive model is trained on thousands of images and their corresponding human recall data.

• Score prediction: The trained model analyzes new assets and predicts a Memory Score based on visual features that correlate with human recall performance.

• Video analysis: For videos, frame-by-frame analysis is performed to produce overall and time-based memory scores.

The resulting score reflects the likelihood of passive viewers remembering the asset after brief exposure (less than 2 seconds) and subsequent distraction, combining both explicit recall accuracy and implicit reaction time measures.

 

  Engagement score

The Engagement score represents the positive emotional response to an image or video asset based on viewers' gut reactions. It measures the intensity of positive emotions like interest, engagement, and happiness that the asset evokes in the audience.

How do I interpret the scores?

• Low scores (0-24) indicate that the asset fails to generate a strong positive emotional response. Viewers may find the content uninteresting or not engaging.
• Medium scores (25-74) suggest that the asset elicits a moderate positive emotional response. There's room for improvement, but the content has some appeal.
• High scores (75-100) mean the asset is highly effective in generating positive emotions. Viewers are likely to find the content engaging, interesting, and potentially happiness-inducing.

How is the score calculated?

The Engagement score is calculated based on our a sophisticated FRT (fast response test) methodology:

1. Participants are shown the asset (image or video) for a brief period (about 5 seconds).
2. They are then presented with association words like "Engaging," "Interesting," and "Happy."
3. Participants respond quickly with "yes" or "no" to these associations.
4. The score takes into account:
   • The yes/no responses
   • Response times (faster responses indicate stronger associations)
   • Group consensus among participants
5. For videos, the score is calculated frame by frame and then aggregated.
6. The final score is determined using advanced machine learning models trained on extensive datasets, correlating visual features with human responses.

This method provides a robust measure of the asset's ability to evoke positive emotional engagement, based on immediate gut reactions rather than prolonged consideration. 

Our deep learning model is then trained to predict the scores based on extensive human-based studies. 

 

---

 

Legacy metrics

  Clarity score

The Clarity score predicts whether customers will experience your asset clearly or not.

How do I interpret Clarity scores?

A low Clarity score means that customers will experience your design cluttered and hard to understand. Designs with much information, or even noise, typically lead to lower Clarity scores. Here, you should ask whether the many and/or noisy elements are important in conveying the message in your design.

In most instances, Clarity should be high, combined with high attention to your main assets (brand, offer, message, product). Under certain circumstances, a low Clarity is acceptable, as in when your asset is showing an abundance of offers, and there is no single item that should grab attention.

 

  Engagement score

The Engagement score is the prediction of how excited and immersed your customers will feel when they are exposed to your asset. A high level of Engagement leads to increased brand memory and purchasing behavior.

How do I interpret Engagement scores?

We often see that Engagement is driven by emotional elements, such as faces showing emotions, as well as objects, large letters, and offers. Generally clear and simple visual information.

If you have a strong brand, consider making the brand an important part of your design, as this will lead to higher Engagement.

We often see that lower Engagement is driven by factors such as complex visual designs, gloomy colors, and highly technical materials. To increase Engagement, consider simplifying your messaging; less is more!