1561 | Change image from flat to model in shopping cart | Mobile
Last updated:30 October 2026
First validation:04 Mar 2026
Last validation:04 Mar 2026
Number:1561
A/B test results
Short description
In this experiment, we tested replacing flat product images with model photography in the mobile shopping cart.
Tips for applying to your e-commerce site
Do’s
Use clear, product-focused images in the shopping cart.
Prioritize quick product recognition and easy order verification.
Use model photography on product detail pages, where customers are still evaluating products.
Don'ts
Assume that visually engaging images always improve conversions.
Introduce unnecessary visual distractions during checkout.
This hypothesis is grounded in these main psychological principles:
Reduce Uncertainty: Showing products on models helps customers understand their fit, proportions, and appearance.
Motivation: Model photography makes products more relatable and appealing by helping customers imagine themselves wearing them.
Based on this background, we believe that replacing flat product images with model photography in the shopping cart for mobile shoppers will cause increased purchase confidence and higher conversion rates. We will know this when we see online transactions increase.
Across both experiments, replacing flat product images with model photography did not improve conversions. Mobile showed a negative effect, while desktop remained unaffected.
This suggests that customers may prefer clear product recognition over additional visual inspiration at the shopping cart stage.
Possible explanations for why the impact was not as hypothesized:
Flat product images could make it easier for customers to quickly recognize and verify the exact items they selected.
At the cart stage, customers may prioritize checking their selected products rather than receiving further inspiration.
Reasons for differences between devices:
Mobile shoppers may rely more heavily on product thumbnails to quickly verify their selections.
Desktop shoppers can more easily review product information alongside images, reducing the influence of image style.
General
All A/B tests in Evidoo have been analysed using Bayesian Statistics. The most important advantage of Bayesian statistics is that it is easy to understand. If there is a difference between the control and the variant, we determine the probability that there is a difference. The probability that the variation differs from the control, is indicated in a percentage.
An A/B test labeled as ">80%" (winner), indicates that the hypothesis has a high probability (>80%) of being true.
An A/B test labeled as "21 - 79%" (inconclusive), suggests the hypothesis has an intermediate chance of being true. This probability range indicates that there is still uncertainty regarding the hypothesis. Therefore it can not be clearly categorized as true or false.
An A/B test labeled as "< 20%" (loser), likely represents a hypothesis that has less than a 20% chance of being true. This suggests that the hypothesis is likely false.
If the primairy KPI is ‘transactions’, there is no impact on average order value, unless mentioned in the learnings.
All A/B tests in Evidoo are also analysed on secondary KPIs, for example 'add to carts'. If we found remarkable results on other KPI’s, check the tab 'learnings'. We also analysed different segments, for example 'new visitors' or 'returning visitors'. If we found remarkable results on specific segments, check also the tab 'learnings'.
All conducted A/B tests in Evidoo comply with:
- conducting sample size and power checks.
- performing the experiment only after an A/A test has been completed.
- implementing Sample Ratio Mismatch (SRM) checks.
- maintaining a minimum runtime of 2 weeks.
- measuring A/B tests only when they are visible to the website visitor, such as counting an A/B test in the middle of the product page only if the visitor has scrolled to that point.
All A/B tests in Evidoo have been analysed using Bayesian Statistics. The most important advantage of Bayesian statistics is that it is easy to understand. If there is a difference between the control and the variant, we determine the probability that there is a difference. The probability that the variation differs from the control, is indicated in a percentage.
An A/B test labeled as ">80%" (winner), indicates that the hypothesis has a high probability (>80%) of being true.
An A/B test labeled as "21 - 79%" (inconclusive), suggests the hypothesis has an intermediate chance of being true. This probability range indicates that there is still uncertainty regarding the hypothesis. Therefore it can not be clearly categorized as true or false.
An A/B test labeled as "< 20%" (loser), likely represents a hypothesis that has less than a 20% chance of being true. This suggests that the hypothesis is likely false.
If the primairy KPI is ‘transactions’, there is no impact on average order value, unless mentioned in the learnings.
All A/B tests in Evidoo are also analysed on secondary KPIs, for example 'add to carts'. If we found remarkable results on other KPI’s, check the tab 'learnings'. We also analysed different segments, for example 'new visitors' or 'returning visitors'. If we found remarkable results on specific segments, check also the tab 'learnings'.
All conducted A/B tests in Evidoo comply with:
- conducting sample size and power checks.
- performing the experiment only after an A/A test has been completed.
- implementing Sample Ratio Mismatch (SRM) checks.
- maintaining a minimum runtime of 2 weeks.
- measuring A/B tests only when they are visible to the website visitor, such as counting an A/B test in the middle of the product page only if the visitor has scrolled to that point.
Questions? Visit our FAQ page. Need any advice?
Contact us
