2. Experiment 1
Retailer & Products
Retailer of drink accessories (glasses, decanters, etc).
Hypothesis:
Product recommendations on the product page, showing similar
alternatives to the product being viewed, will help visitors find what
they’re looking for, and thus increase the conversion rate.
Solution:
Product recommendations on product page, below main product image.
3. Variation A - 6x1 recommendations, algorithm-driven
Variation B - 5x1 recommendations, algorithm-driven
Control - no recommendations
4.
5.
6. Experiment 2
Hypothesis:
Product recommendations on the cart page, showing complementary
additions to the product(s) in the shopper’s cart, will increase average
order value.
Solution:
Product recommendations on cart page, below shopping cart details and
‘checkout’ button.
7. Variation A - 6x1 recommendations, algorithm-driven
Variation B - 5x1 recommendations, algorithm-driven
Control - no recommendations
Variation A - 6x1 recommendations, algorithm-driven
Variation B - 5x1 recommendations, algorithm-driven
8.
9.
10. 7 things we’ve learnt at Bunting that will
improve your recommendation performance
11. 7 things we’ve learnt at Bunting that will
improve your recommendation performance
1) Quantity / image size can dramatically affect performance
12. 7 things we’ve learnt at Bunting that will
improve your recommendation performance
1) Quantity / image size can dramatically affect performance
2) Use isolated images, rather than lifestyle images
13. 7 things we’ve learnt at Bunting that will
improve your recommendation performance
1) Quantity / image size can dramatically affect performance
2) Use isolated images, rather than lifestyle images
3) Place your recommendations directly below main product photo /
summary description, but above product description
14. 7 things we’ve learnt at Bunting that will
improve your recommendation performance
1) Quantity / image size can dramatically affect performance
2) Use isolated images, rather than lifestyle images
3) Place your recommendations directly below main product photo /
summary description, but above product description
4) Recommendations grids: horizontal generally perform better on
desktops, while vertical should be used on small mobile devices
15. 7 things we’ve learnt at Bunting that will
improve your recommendation performance
1) Quantity / image size can dramatically affect performance
2) Use isolated images, rather than lifestyle images
3) Place your recommendations directly below main product photo /
summary description, but above product description
4) Recommendations grids: horizontal generally perform better on
desktops, while vertical should be used on small mobile devices
5) Remember the context! Product page recommendations should be
similar alternatives to the main product. Post-add-to-cart
recommendations should be complimentary extras to the cart
contents
16. 7 things we’ve learnt at Bunting that will
improve your recommendation performance
1) Quantity / image size can dramatically affect performance
2) Use isolated images, rather than lifestyle images
3) Place your recommendations directly below main product photo /
summary description, but above product description
4) Recommendations grids: horizontal generally perform better on
desktops, while vertical should be used on small mobile devices
5) Remember the context! Product page recommendations should be
similar alternatives to the main product. Post-add-to-cart
recommendations should be complimentary extras to the cart contents
6) Use big data recommendations, rather than human assigned
17. 7 things we’ve learnt at Bunting that will
improve your recommendation performance
1) Quantity / image size can dramatically affect performance
2) Use isolated images, rather than lifestyle images
3) Place your recommendations directly below main product photo /
summary description, but above product description
4) Recommendations grids: horizontal generally perform better on
desktops, while vertical should be used on small mobile devices
5) Remember the context! Product page recommendations should be
similar alternatives to the main product. Post-add-to-cart
recommendations should be complimentary extras to the cart contents
6) Use big data recommendations, rather than human assigned
7) Split test different recommendation content
(eg: most popular vs. related)
Variables influencing the result
Size of product recommendation images
Smaller recommendation images of Variation A will reduce the detail, which could affect the incentive to click.
Number of recommendations
Recommendations 1-5 are identical in both variations. However the additional 6th recommendation in Variation A will raise the chance of showing the visitor the right desirable alternative for them.
Vertical height
Taller images on Variation B could increase the chance of that image / product details are cropped below the fold.
Variables influencing the result
Size of product recommendation images
Smaller recommendation images of Variation A will reduce the detail, which could affect the incentive to click.
Number of recommendations
Recommendations 1-5 are identical in both variations. However the additional 6th recommendation in Variation A will raise the chance of showing the visitor the right desirable alternative for them.
Vertical height
Taller images on Variation B could increase the chance of that image / product details are cropped below the fold.
sdsd