A character's gaze steered visitors' attention.
DIRECTION OF GAZE
Changed the distribution of attention
104
respondents in the experiment
ηp² = 0.11
medium-sized effect · p < 0.001
01 — WHAT THE DATA SHOWED
01
Changing the direction of the character's gaze had a statistically significant effect on the distribution of visual attention: F(1,102) = 12.57; p < 0.001.
02
The effect size ηp² = 0.11 corresponds, by commonly used rule-of-thumb thresholds, to a medium-sized effect.
03
In the tested creatives the character's composition thus worked as a visual cue. The study, however, did not measure clicks, message recall or conversion.
02 — WHAT WE NEEDED TO FIND OUT
Does the direction of the character's gaze change the distribution of attention in a creative?
A human face in advertising attracts attention. The experiment tested whether changing the direction of the character's gaze also affects how respondents distribute their attention between the face and the other parts of the tested creative.
03 — HOW THE MEASUREMENT WORKED
104 respondents took part in the eye-tracking experiment. The creative variants differed in the direction of the character's gaze, and the analysis compared the distribution of visual attention between predefined areas.
[ PRINCIPLE OF THE TESTED VARIANTS ]
The diagram only explains the manipulated variable. It does not show the actual stimuli or the participants' heatmap; the statistics come from a real measurement.
05 — RECOMMENDATION FOR PRACTICE
Include gaze direction among the variables when designing and testing creatives. If the character is meant to support attention to a specific message or product, prepare controlled variants and validate them together with campaign metrics.
[ THE TAKEAWAY ]
Even a small change of composition can shift the distribution of attention.
Eye tracking helps choose a variant based on what people actually notice. The impact on clicks or conversion needs to be verified with follow-up data.
DISCUSS A CREATIVE TEST →[ GLOSSARY — WHAT THE TERMS IN THIS STUDY MEAN ]
- N (sample)
- The number of people included in the analysis. In this study N = 104.
- F(1,102)
- The value of the test statistic and its degrees of freedom in the statistical test used.
- p-value
- Helps assess how compatible the data are with the assumption that gaze direction had no effect in the model used.
- ηp²
- Partial eta squared expresses the effect size in the given statistical model. It is not a percentage increase in campaign performance.
[ FREQUENTLY ASKED QUESTIONS ABOUT THE STUDY ]
Did the direction of the character's gaze affect attention?
Yes; in the tested creatives and on a sample of 104 respondents, the distribution of attention between the variants differed to a statistically significant degree: p < 0.001; ηp² = 0.11.
Does this mean we should always turn the character toward the CTA?
Not automatically. The result supports gaze direction as a relevant compositional variable. The suitable variant depends on the specific photo, message, format and goal of the creative, so we recommend comparing real designs.
Did the study prove more clicks or conversions?
No. The study measured the distribution of visual attention, not clicks, recall, purchase or any other business metric. A shift of attention is an important intermediate step, but on its own it does not guarantee action.
What does ηp² = 0.11 mean?
Partial eta squared describes the effect size in the statistical model used. By common rule-of-thumb thresholds the value 0.11 is interpreted as a medium-sized effect; it does not, however, automatically mean an 11% increase in campaign performance.
What does p < 0.001 mean?
If gaze direction in fact had no effect in the model used and its assumptions were met, a result this pronounced or more would be very unusual. The p-value itself does not describe the size or the business value of the effect.
Do we have to redo the whole creative for a test?
Not necessarily. Often it is enough to prepare controlled variants of the photo, crop or position of the character. The other elements should stay as similar as possible so the difference can be interpreted meaningfully.
Want to go through the study in more detail?
DETAILED REPORT WITH METHODOLOGY AND RECOMMENDATIONS — FREE PDF