Packaging design has to communicate several things at once. A coffee package, for example, may need to establish the product and origin, communicate flavor, indicate roast level and coffee type, reinforce the brand, and still remain visually appealing.
But not every element should receive the same amount of attention.
In this case study, we used Attention Insight’s predictive attention analysis to conduct packaging A/B testing as part of a pre-launch creative optimization study. The results reflect improvements in predicted visual performance rather than measured increases in click-through rates (CTR), registrations, or conversions.
Rather than asking which design is universally “better,” the test began by defining a specific attention goal. The two variants were then compared to determine which visual hierarchy better supported that goal.
Defining the Goal Before A/B Testing
Before conducting packaging A/B testing, it is important to decide what you actually want people to notice.
There is no single ideal hierarchy for every coffee package. One brand might prioritize its name, while another may want shoppers to notice origin, roast level, tasting notes, or whether the coffee is whole bean or ground.
The goal therefore needs to be established before interpreting the attention results.
For this case study, Tasting Notes were selected as the main attention target. Product Name / Origin should remain the primary identifier, while Roast Level and Coffee Type should remain visible as useful secondary information.
The intended hierarchy was:
Product Name / Origin → Tasting Notes → Roast Level & Coffee Type → Features → supporting brand and visual elements
Hypothesis
A coffee packaging variant that gives greater visual prominence to Tasting Notes will attract more attention to this information while maintaining Product Name / Origin as the primary focal point and preserving the visibility of Roast Level and Coffee Type.
Having a predefined goal gives us something concrete against which to evaluate the results rather than simply looking for the design with the highest attention percentages.
Comparing Two Coffee Packaging Variants
For this predictive packaging A/B test, we compared two different packaging variants for the same coffee product.
Both contain the same essential product information and maintain a consistent visual identity, but they use noticeably different layouts and visual hierarchies to communicate that information.
Rather than treating Design A as a “bad” package and Design B as its corrected version, the two designs represent different packaging approaches that could realistically be used for the same product.
Design A
In Design A, the product information panel occupies the upper part of the package, while the coffee illustration extends across the lower portion.
Product Name / Origin is given substantial visual prominence. Features such as varietal, elevation, and processing also occupy a relatively prominent position within the information panel.
Tasting Notes, Roast Level, and Coffee Type are comparatively understated.
Design B
Design B contains the same core information but organizes it differently.
The illustration occupies the upper portion of the package, while the structured product-information panel sits below it. Most importantly, Tasting Notes are presented in a wide contrasting section, using a warm orange background and substantially larger flavor descriptors.
Product Name / Origin remains prominent above it, while Features, Roast Level, and Coffee Type form supporting information beneath.
The question is therefore not simply which layout looks better. It is:
Which of these visual hierarchies better supports the goal established for this particular packaging A/B test?
Setting Up the Areas of Interest
Using Attention Insight, we generated predictive attention heatmaps and defined Areas of Interest (AOIs) to measure the predicted attention received by each key packaging element.
The same information categories were marked as AOIs in both designs:
Product Name / Origin, Tasting Notes, Features, Roast Level, Coffee Type, Image, and Label.
Where the layout differed between designs, the AOIs were drawn around equivalent functional areas rather than attempting to make the boxes physically identical.
For example, the Tasting Notes AOI represents the complete dedicated Tasting Notes area in each design. In Design B, this includes the larger contrasting section containing the heading, flavor descriptors, and associated visual treatment.
This allows the comparison to focus on the same semantic information across two different layouts.
Design A: Predictive Attention Analysis
The predictive attention analysis revealed a clear hierarchy in Design A.
| Area of Interest | Design A |
|---|---|
| Product Name / Origin | 25% |
| Image | 14% |
| Features | 13% |
| Tasting Notes | 2% |
| Coffee Type | 1% |
| Roast Level | 1% |
| Label | 0% |
Product Name / Origin was the strongest information area, receiving approximately 25% of predicted attention.
However, the next strongest AOIs were the Image at 14% and Features at 13%.
The Tasting Notes – the information selected as the main target for this test – received only 2%.
Roast Level and Coffee Type each received approximately 1%, while the Label did not register measurable attention in the rounded AOI comparison.
For our chosen goal, the main issue is therefore clear – Tasting Notes do not form a meaningful focal point in Design A.
This does not make Design A objectively unsuccessful. If the testing goal were to emphasize Product Name / Origin, Features, or illustration, the same results could be interpreted differently.
But against the hierarchy defined for this particular test, there is considerable room for improvement.
Design B – A Different Visual Hierarchy
Design B produced a substantially different predicted attention distribution.
| Area of Interest | Design A | Design B | Change |
|---|---|---|---|
| Product Name / Origin | 25% | 27% | +2 pp |
| Tasting Notes | 2% | 19% | +17 pp |
| Features | 13% | 10% | −3 pp |
| Image | 14% | 3% | −11 pp |
| Coffee Type | 1% | 3% | +2 pp |
| Roast Level | 1% | 2% | +1 pp |
| Label | 0% | 1% | +1 pp |
Values are displayed as rounded percentages in the Attention Insight comparison.
The results reveal a considerable redistribution of attention.
Tasting Notes – 2% → 19%
The most important result concerns our primary testing target.
Predicted attention to Tasting Notes increased from approximately 2% in Design A to 19% in Design B – an increase of 17 percentage points.
Tasting Notes therefore went from being one of the least noticeable pieces of product information to becoming the second-strongest measured AOI.
The visual treatment in Design B – particularly the dedicated contrasting section, increased typography, and clearer separation from surrounding information – created a substantially stronger focal point for the flavor profile.
Product Name / Origin – 25% → 27%
Importantly, increasing attention to Tasting Notes did not require sacrificing the primary identifier.
Product Name / Origin remained the strongest AOI and increased from approximately 25% to 27% in the flat-design comparison.
The resulting hierarchy therefore corresponds closely with our original goal:
Product Name / Origin → Tasting Notes → supporting product information
This supports the hypothesis established before testing.
Attention Was Redistributed, Not Simply Increased
One of the most useful findings from this comparison is what happened to the other elements.
Features decreased from 13% to 10%, while the Image decreased from 14% to 3%.
At first glance, lower attention percentages can appear negative. But attention should not be maximized across every element.
In Design A, the illustration received approximately seven times as much predicted attention as Tasting Notes, while Features received more than six times as much.
In Design B, that relationship changed substantially.
Attention shifted away from supporting and decorative content and toward the information selected as more important for this particular communication goal.
This demonstrates an important principle of predictive packaging testing: more attention is not automatically better. What matters is whether attention is distributed according to the intended visual hierarchy.
If every element is enlarged, made brighter, or given stronger contrast, the package may simply become more visually competitive.
Effective hierarchy requires some elements to lead and others to support them.
Secondary Information Also Became More Visible
Coffee Type received a slightly higher share of predicted attention in Design B, increasing from approximately 1% to 3%, while Roast Level remained a relatively low-attention element.
These secondary details were not the primary focus of the test, and their small changes should not be overinterpreted. The more important finding is that the overall hierarchy shifted toward Tasting Notes without displacing Product Name / Origin as the dominant information area.
Comparing the Final Hierarchies
Comparing the Final Hierarchies
The side-by-side comparison highlights how the two layouts prioritize information differently. Design A gives greater visual emphasis to the illustration and technical product details, while Design B brings the flavor profile forward through a larger, contrasting Tasting Notes section.
This distinction illustrates how design decisions can redistribute predicted attention without changing the information itself.
What the Focus Maps Reveal
The Focus Maps provide another way of seeing the difference between the two packaging variants.
In Design A, visibility is concentrated strongly around Product Name / Origin and parts of the Features section, while Tasting Notes have little visual presence.
In Design B, Product Name / Origin remains highly visible, but the Tasting Notes section also becomes a clearly visible secondary focal area.
This reinforces what the AOI percentages show quantitatively – Design B creates a more deliberate two-level hierarchy between product identification and flavor information.
Which Packaging Variant Better Supports the Goal?
For the predefined objective, Design B is the stronger variant. However, this conclusion is specific to the communication goal established before testing.
A designer prioritizing imagery or technical coffee information might interpret the same comparison differently. This is why predictive attention results should be evaluated against a clear objective rather than treated as a universal design score.
Applying the Designs to the Actual Packaging
The initial comparison was performed using the flat packaging artwork. This makes it easier to examine visual hierarchy without additional variables introduced by the physical presentation of the package.
The next step was to apply both designs to their intended coffee bag format and test them again as pack shots.
This second stage of packaging A/B testing provides an additional check of whether the intended attention hierarchy remains visible once the artwork is presented on a more realistic package.
Testing Attention on the Pack Shots
The pack-shot comparison produced the following results:
| Area of Interest | Pack Shot A | Pack Shot B | Change |
|---|---|---|---|
| Product Name / Origin | 27% | 25% | −2 pp |
| Tasting Notes | 1% | 18% | +17 pp |
| Features | 9% | 8% | −1 pp |
| Image | 18% | 13% | −6 pp |
| Coffee Type | 1% | 2% | +1 pp |
| Roast Level | 1% | 2% | +1 pp |
| Label | 1% | 4% | +3 pp |
Values are displayed as rounded percentages in the Attention Insight pack-shot comparison.
The physical presentation changed some of the individual percentages, but the central finding of the original A/B test remained highly consistent.
Most importantly, Tasting Notes increased from 1% in Pack Shot A to 18% in Pack Shot B – again a 17 percentage-point increase.
This closely mirrors the flat-design result of 2% → 19% (+17 pp).
The Main Hierarchy Was Preserved
Product Name / Origin changed from approximately 27% to 25% in the pack-shot comparison. Unlike the flat-design test, where it increased slightly, it therefore decreased by around 2 percentage points.
However, this does not contradict the original hypothesis.
The goal was not necessarily to increase attention to Product Name / Origin, but to maintain it as the primary focal point while substantially increasing attention to Tasting Notes.
That hierarchy remained intact.
In Pack Shot B, Product Name / Origin received approximately 25% of predicted attention, while Tasting Notes received 18%.
The two dominant information areas were therefore still:
Product Name / Origin → Tasting Notes
This is consistent with the intended hierarchy established before testing.
Flat Artwork vs. Pack-Shot Results
Looking at both stages together provides a useful comparison:
| Area of Interest | Flat A | Flat B | Pack A | Pack B |
|---|---|---|---|---|
| Product Name / Origin | 25% | 27% | 27% | 25% |
| Tasting Notes | 2% | 19% | 1% | 18% |
| Features | 13% | 10% | 9% | 8% |
| Image | 14% | 3% | 18% | 13% |
| Coffee Type | 1% | 3% | 1% | 2% |
| Roast Level | 1% | 2% | 1% | 2% |
| Label | 0% | 1% | 1% | 4% |
The results should not be expected to remain identical once the artwork is placed on a physical package. Package shape, presentation, visible surface area, and the relationship between the artwork and the bag itself can all affect the predicted distribution of attention.
The Image shows the clearest contextual difference. In the flat comparison, Image attention decreased from 14% to 3%, whereas in the pack-shot comparison it decreased from 18% to 13%.
The illustration therefore retained substantially more attention when presented on the physical bag.
The Label also became more prominent in the pack-shot comparison, increasing from 1% to 4% in Design B.
These differences are useful rather than problematic. They demonstrate why a pack-shot test can complement the controlled flat-design analysis by revealing how the same hierarchy behaves when additional physical context is introduced.
Most importantly, the central result remained remarkably stable:
Flat artwork – Tasting Notes: 2% → 19% (+17 pp)
Pack shot – Tasting Notes: 1% → 18% (+17 pp)
In both contexts, Tasting Notes changed from a very low-attention element into the second-strongest information AOI, while Product Name / Origin remained the primary focal point.
The pack-shot test therefore provides additional support for the conclusion drawn from the flat-design comparison rather than replacing it.
Taking Packaging Testing Further – Shelf Context
Packaging rarely appears entirely on its own.
In a real shopping environment, a coffee bag may compete with many other packages for attention. That introduces a different question:
Does the package itself stand out sufficiently among competing products?
After comparing visual hierarchy within individual packaging variants, the selected design can therefore also be tested within a shelf or retail context.
This creates two complementary levels of analysis:
Individual package test – Where is attention distributed within the package?
Shelf-context test – Does the package itself attract attention among competing products?
Together, these tests can provide a broader understanding of packaging visibility before launch.
What This Packaging A/B Test Shows
This case study demonstrates how predictive attention testing can help designers evaluate packaging decisions against a predefined communication goal.
Using Attention Insight, we found that Design B increased predicted attention to Tasting Notes by approximately 17 percentage points in both the flat-artwork and pack-shot comparisons, while Product Name / Origin remained the strongest information area.
The consistency across both testing contexts suggests that the revised visual hierarchy is not limited to the flat artwork. It also remains effective when applied to a more realistic coffee package presentation.
However, the strongest design is not necessarily the one that attracts the most attention to every element. A different brand or product might prioritize origin, roast level, brand recognition, or another piece of information entirely.
The key is to establish what shoppers should notice first, then use predictive attention analysis to compare how well different designs support that objective.
By testing both the artwork and its application to packaging, designers can make more informed decisions about visual hierarchy before moving into further validation or production.
