Case Study
Sun & Ski Sports Optimizes Product Recommendations
Founded
1980
Industry
Outdoor Gear & Clothing
Headquarters
Houston, Texas
Retail Locations
30 Store Locations
About Sun & Ski Sports
One of the largest specialty outdoor retailers in America, Sun & Ski Sports is inspiring you to let adventure into your life confidently through exceptional customer service, first-hand expertise, and competitively priced brands ready to perform for any journey – to be where you live and play. Their trademark experience has placed the company at the top of specialty outdoor retailers in America.
Sun & Ski Sports has grown to include 30 stores in 12 states across the country, including sunandski.com, yet still manages to provide that small store feel with big store competitive pricing. They specialize in equipment, apparel, footwear, and accessories to enhance their customer’s active lifestyles and year-round activities including ski, snowboard, bike, run, swim, and more. They deliver a small store feel with big store competitive pricing with a vision to be where their customers live and play. Their unique model is based on localization, specializing in regional inventory depending on the climate and popular activities, as well as a thriving rental business.
The Digital Merchandising team at Sun & Ski Sports set out to optimize their product recommendations in two phases: First, they compared SiteSpect product recommendations against their existing product recommendations provider. Next, they compared two variations of SiteSpect product recommendations.
Optimizing Product Recommendation Placement
Once Sun & Ski Sports signed on with SiteSpect, they gained the ability to test and optimize their existing product recommendations platform. Because SiteSpect can integrate with virtually any other tool, they were able to track their existing product recommendations provider metrics side-by-side with their SiteSpect product recommendations metrics. Their first step was to identify the impact of where they placed recommendations, using both SiteSpect and their existing product recommendations provider. This allowed the team to learn about the relative effectiveness of different recommendation placements and refine the recommendation strategy.
The brand compared the following placements and algorithms:
- Home Page, Personalized
- Category Pages, Personalized
- Add to Cart Modal, Complementary
- Cart Page, Complementary
- Product Detail Page, People Also Like
This test ran during Black Friday and Cyber Week, which also saw an increase from normal traffic levels. This allowed Sun & Ski Sports to release the test to 20% of traffic, with that percentage split evenly between SiteSpect and their existing product recommendations provider.
Results showed that the Product Detail Page saw 59% of all recommendation clicks, followed by Category pages and the Add to Cart Modal.
Comparing Product Recommendations Strategies: SiteSpect vs. Existing Product Recommendations
With the knowledge gained from their first A/B test, Sun & Ski Sports compared two different strategies of SiteSpect product recommendations alongside their existing product recommendations in an A/B/C test on Product Detail Pages. The recommendations were all based on a “People Also Like” algorithm. Each SiteSpect variation used different boosting rules.
“People Also Like” product recommendations are based on a variety of factors that make up the “signature” of a given product. The signature is made up of behaviors and other predictors, including searches, product views, and category views that lead to purchasing an item. The model leverages these behaviors to predict items that show similar purchase behaviors, and lead to higher conversion rates.
Variation 1 of this test applied SiteSpect product recommendations using this algorithm, and then also boosted items from the same brand, gender, and category. This means that these boosted items were prioritized and shown first and more frequently in the recommendations module. Variation 2 also used SiteSpect product recommendations and the same algorithm but did not boost items with the same brand. Sun & Ski Sports hypothesized that removing brand filtering would open up more cross-sell options and help users find and compare relevant products more efficiently. Finally, Variation 3 used their existing product recommendations with their comparable “People Also Like” algorithm and no boosting.
After running the experiment for 16 days, the SiteSpect recommendations outperformed compared to their existing product recommendations by 5.8% based on Revenue per Visit. Those results translated in net proceeds of over $117K just from this test.
Other findings showed that Variation 1, which boosted recommendations by brand, led to more purchases but slightly smaller average order value than Variation 2. Both boosting strategies led to statistically similar Revenue per Visit.
Optimizing Product Recommendations Going Forward
In response to these promising results, Sun & Ski Sports is taking steps to further optimize recommendations on their site. As they move forward with SiteSpect, they will load the historical data they’ve acquired over the past two years to hone the seasonality of their recommendations. They will further test and analyze behavior to optimize recommendations on mobile, as well as fine-tune recommendations across their site.
