Puma increases checkout revenue by +21.66% with real-time customer experience optimization

Puma story.

In 1924, Rudolf and his brother Adolf “Adi” Dassler had jointly formed the company Gebrüder Dassler Schuhfabrik (Dassler Brothers Shoe Factory). The relationship between the two brothers deteriorated until the two agreed to split in 1948, forming two separate entities, Adidas and Puma.

Following the split from his brother, Rudolf originally registered the newly established company as Ruda (derived from Rudolf Dassler, as Adidas was based on Adi Dassler), but later changed the name to Puma.

The official Puma online store sells the world’s largest selection of Puma shoes, clothing, accessories, and equipment. These products typically fall into one of two categories: (1) sports performance and lifestyle, and (2) sports fashion. 

Puma’s performance label includes categories such as soccer, running, motorsport, and more. Their fashion line, on the other hand, features collaboration with renowned designers such as Alexander McQueen, Sergio Rossi, and many others. 

Famous for fashionable performance footwear, Puma has created many popular sneaker styles, including the Suede, El Ray, Roma, Speed Cat, and Cabana Racer. The brand’s motorsport line also includes officially licensed logo wear from Ferrari and Ducati.

Before

Obviyo was introduced to Puma by our agency partner Netrada, as an experience optimization platform with unique algorithmic capability to experiment with many variables in parallel to produce a sustainable lift in RPV (revenue per visitor) in real time. 

While the ultimate goal was to optimize the full end-to-end user experience on the Puma’s e-commerce site, the checkout funnel optimization was identified as the initial priority. According to Shopify 69.75% of desktop shoppers and 85.65% of mobile shoppers abandon their carts.

After

The scope of the optimization campaign included the multivariate optimization of Puma US’s checkout funnel. Netrada UX team created new checkout funnel designs which are used as the input for formulation of experiment hypothesis. Key new design ideas are turned into optimization campaign variables. Obviyo provisioned these multivariate changes that spanned across the entire checkout funnel. This approach ensured end-to-end results and a consistent user experience throughout checkout.

The optimization campaign included the following content and functional variations:

A – Main Navigation – suppresses the main navigation area including the search option

B – Cart Size/Emphasis – increases the font size in the funnel and make button call to action bold

C – Trust Badge – move the Norton badge up higher on the page throughout the entire checkout funnel

D – Coupon – present the promo/coupon code form field up higher on the page

E – Checkout – bypasses the Login page to eliminate a step during checkout process

H – Form Fields, Wishlist update – on the cart page replace Add to Wishlist with Update option; add highlight to form fields in funnel while eliminating grey background.

I – Layout – suppresses the redundant credit card options in checkout; changes layout of payment options to have Puma Gift Card on right

J – Newsletter – on order review page puts Puma Newsletter signup at bottom left of page and spreads the order summary info across the page.

The main conversion goal was a successful completion of the checkout while optimization goal was to maximize RPV (revenue-per-visitor).

Results

Use of adaptive algorithms to experiment with many A/B variables in parallel is accelerating time to results while enabling ecommerce brands to make money even as they are experimenting with many negative treatments.

Namely, the algorithms are dynamically allocating traffic to winning treatments to over compensate for the drag of negative ones.

Table below shows how the entire campaign, all treatment listed above playing as a group, produced +21.66% RPV lift during the time interval during which this campaign was running.

Variable Importance

The table below shows how individual experiment treatments performed on average. An uneven distribution of visitors across different variables illustrates how algorithms are dynamically allocating more traffic to winning treatments and much less on the negative ones.

Top ensembles

There were 256 permutations of variables included in the campaign, which we call ensembles.

Similar to not evenly splitting traffic among participating experiment variables, the algorithms are not evenly splitting traffic among ensembles either. Instead, the ensembles that perform better are getting a higher percentage of the overall traffic.