In this experiment, a feint gray background search bar was replaced with a higher contrast one containing a blue border and a filled blue color button style.  Impact on searches and purchases was measured. (Inverted to fit the Filled-vs-ghost button pattern).

In this experiment, popular products were shown at the top of the homepage. Impact to sale was measured.

In this experiment, the variation transformed an add-to-cart button into a shopping cart one (making it a two step process). The control had an interaction where an add-to-cart button would become a quantity selection and then a shopping cart action (a three step process). In a way, the control kept users in a "dead-end" quantity selection middle state that required clicking the back button to continue the sale. Whereas the variation automatically moved users into the next step (closer towards being able to make a purchase). Impact on sales was measured. 

In this experiment, recently purchased products were appended at the top of the homepage. The test ran for loggedin users only. Impact on add-to-cart, sales and revenue was measured.

In this experiment, customer reviews were shown on product detail pages. Impact on adds to cart and sales was measured.

Are more (popular) product results better than none at all? In this experiment, popular products were shown during an empty search result. Impact on sales was measured.

In this experiment, popular product category links were added near the bottom of the homepage of an online grocery store - Metro. Impact on completed sales was measured.

In this experiment, popular categories were added at the bottom of a long ecommerce homepage. Impact on total sales was measured.

In this experiment, the interaction on a product page was changed. In the control (A), after clicking "add to cart", the customer would remain on the product page with an updated basket and the button changing to "Go To Cart". The variation (B) however took customers forward automatically to the cart.

In this experiment, reviews on product pages were exposed completely. Instead of hiding them under a clickable tab, they were shown by default in the variation. Impact on sales was measured.

In this experiment, popular search terms were added at the bottom of a long e-commerce homepage. Hence, the variation showed additional search triggers that lead to results pages. (Translated from Russian using Google Translate)

In this experiment, the discounted price was shown along with an active promotion. The control only showed that the relative -8% discount was present with the pre-discounted price.

In this experiment, products which were out of stock were removed from listing pages and replaced with in stock ones (not visible in the screenshot).