Dragonfly AI helps you to maximise the performance of physical and digital content and experiences by showing you instantly what your audience will see first. Dragonfly AI gives you predictive analytics while you iterate so you can see what really works before you publish.


Predictive attention analytics fit into any existing design or optimisation workflow and can be used to optimise content pre and post-production.


Setting clear objectives is critical to getting the most out of the optimisation process. Objectives and Key Results (OKR) is a good framework for defining and tracking objectives and their outcomes. Before analysing any content with Dragonfly it’s important that your optimisation priorities are clear.


Objective: Increase percentage of Click & Collect purchases.

Key results

  1. Increase the Click & Collect banner Click Through Rate (CTR) by 3%
  2. Reduce avg. time to complete C&C form by 20%

Attention optimisation can help with a wide range of objectives from simplifying task completion through to improve reach and conversion.


Good content preparation is key to maximising the accuracy and relevancy of the insights. Dragonfly AI tells you what an audience sees first given the content you provide it.


Consider whether the content itself could be variable due to personalisation strategies (e.g. local translations) and whether multiple versions should be tested.


Consider how content will ultimately be viewed once deployed, for example, if content will be viewed online, consider the impact of device or screen size.


Consider how the viewer experience could vary due to changes in context for example an advert seen at night versus during the day or different viewing distances/angles.


Choosing the right attention maps, identifying the most relevant hotspots and creating regions of interest for all your high-value elements provides a great set of data points to make the interpretation phase quick and easy.

Before building your report consider these three key objectives.


Attention maps visualise where the attention of an audience is most and least likely to be directed during first glance.

Heat maps - Use heat to visualise the distribution of attention.

Opacity maps - Use opacity to visualise the distribution of attention.

Analysis is usually carried out in heat mode. Opacity mode is useful when scenes already have strong colours making it difficult to separate heat from the original image.


The scene average score provides an easy way to measure the overall level of busyness within the scene. In the example to the right, the average score is 55/100.


High-value elements in scenes with a low average score have a greater chance of standing out.

This can be measured by looking at the difference between the average attention score for the scene and the region of interest scores for your high-value elements.


Hotspots are the areas most likely to win the attention of your audience during first glance. The brain can only process up to 5 hotspots at any one time, so you’ll want to take a note of the number.

Attention clusters

Use the grid to identify/score hotspots in the scene. If several adjacent cells all score highly, they should be counted at a single cluster/hotspot.

Take note of the average score. If the average is low (e.g. <60) and the hotspots are high (e.g. 80+) this means the hotspots have a good chance of standing out. If the distance between the two numbers is smaller, messaging may struggle to stand out.


Regions allow you to test how specific, high-value elements are performing within the scene (e.g. A call-to-action button on a webpage). Use Dragonfly’s region mode to identify your key regions of interest.

Local Attention Score (LAS)

Your local attention score will help you quantify how attention grabbing your region is. Later, in the experimentation phase, you’ll be able to use these scores to test which versions perform better from an attention perspective.


Interpretation is a critical phase as this your opportunity to transform insights into actions. In this section we’ll guide you through key considerations when interpreting the results in your report.


Before interpreting any analysis it’s important to consider how the viewing context may influence your priorities.

Example: Dwell time

Think about the window of opportunity to deliver your key message. For example, commuters walking past an ad on The London Underground will have a lower dwell time than visitors standing on the platform. In low dwell times contexts, it becomes even more critical to focus attention on only the most important elements.


There are two primary phases to consider during the pre-cognitive phase.

Winning attention

Before you can influence, you need to win the attention of your audience. In the case of product-on-shelf, you would first need to win the attention of a passing shopper before trying to influence them to buy.

Directing attention

Once you have the attention of your audience you need to direct that attention in a way that best aligns with your CX and commercial priorities.


The following are examples of common high-value elements used in design.

Calls to action

Calls to action make it clear to viewers which action to take next and helps remove friction moving down the sales funnel.

Context providers

Users are unlikely to take an action without understanding the context for the action they are taking.


You simply cannot build brand awareness and equity if your branding is never seen.

Emotion impacting elements

Studies show that emotionally charged events create powerful memories in people's minds.


Once you’ve considered the viewing context and target attention strategies, you’ll want to focus on whether the attention distribution aligns with your objectives.

High value elements

Make sure your high-value elements (e.g. calls to action, elements that create an emotional impact and elements which provide a clear context for the next action) are achieving the highest local attention scores.

Rule of 5

The brain has a limited visual attention span and can only process up to 5 distinct visual elements in parallel at any given time so make sure your high-value elements fall within your top 5.


Clarity is an important aspect of effective communication and attention maps can help to identify when too many elements are competing for attention.

Biological limitations

The human brain can only process up to five distinct regions at any given time. Too many elements competing for attention increases the cognitive load on the viewer and reduces your ability to influence what they see first.


The goal of the interpretation phase is to identify one or more hypotheses. A hypothesis is an idea or explanation for something that is based on known facts but has not yet been proven.

Commercial hypotheses

Start by identifying areas where performance could be improved in line with your commercial and customer experience objectives e.g. drawing more attention to a key call to action in order to improve click through rates.

Design hypotheses

Next, consider how specific design changes could help to achieve that outcome e.g. Increasing the size of the call to action button.


Test cards are a great way to plan and record the experiments you’ll be carrying out on your content.

Commercial hypothesis example:

  • We believe that - drawing attention to the summer sale banner will increase sales.
  • To verify that, we will - update the design to draw more attention to the banner.
  • And measure - sales data before and after the change comes into effect.
  • We are right if - sales increase by 1% over 3 months accounting for seasonal variation.

Design hypothesis example

  • We believe that - reducing distractors near the banner will increase its attention score.
  • To verify that, we will - reduce the visual weight of the textured background.
  • And measure - the Local Attention Score of the banner before and after doing so.
  • We are right if - the banner achieves a higher Local Attention Score after the changes.


In the experimentation phase you’ll experiment with design changes to improve the performance of your content. Moving fast and testing frequently is key to getting the best possible results.


Our brains have a limit on how much visual information can be processed. Because of this we have learned to prioritise where to direct our attention based on key visual cues. When experimenting with design changes there are visual characteristics you can play with in order to influence where viewer attention is directed.


If you make everything more attention grabbing nothing will be attention grabbing as the brain will be overloaded.


Imagine you and your team were given 5 votes each and you had to assign those votes to specific elements in the scene that you want the audience to see first. How would you assign those votes?

You may wish to assign all votes to one key message or spread your votes (audience attention) between 5 different elements within the scene. How you assign your votes is subjective based on what you think is most important but should ultimately align to the objectives you set out at the beginning.


When experimenting design changes, there are two key approaches you can take.

Maximising high-value elements

Use the key visual characteristics to make iterative changes to the design until you have the desire attention hierarchy matching your customer experience and commercial objectives.

Minimising distractors

When optimising content the temptation is often to focus all efforts on increasing attention scores for high-value elements but reducing the attention scores for any surrounding, low-value elements can be just as effective.


Running A/B tests can help you to identify which version of your design performs best against your objectives.

Data driven decisions

Measure the performance of both versions using a combination of attention maps, hotspot analysis and regional analysis to see which perform best.


Once you’ve created and tested multiple versions of your design using the approach outlined, you’re ready to pick a winner!

Data-supported decision making

Drawing on a combination of your analysis & interpretation, known constraints (e.g. brand guidelines) and understanding of your target audience now is the time to roll out the new version of your design into the wild.


Once your content is out in the wild, monitor its performance over time to understand the effect of the changes you made. Over time, you’ll build up valuable insights into how attention optimisation can impact performance in different ways. These learnings can then be fed back into your process next time around.

Want to learn more about Dragonfly AI?

For more information on Dragonfly AI please contact [email protected] or any member of the Dragonfly AI team.

Did this answer your question?