Million Dollar Merchandising
How do you know your eCommerce merchandising strategy or technology is effective? Catalogers have tested what products sell and position products strategically on catalog pages with specifically allocated measurements per product. Retailers have optimized their window displays and placement of products on racks since the nineteenth century. eCommerce sites are no different; proper merchandising plays a huge role in their success.
The client for this experiment is a multi-channel retailer with catalog, brick-and-mortar, and web presences, with catalog being their primary channel. They sell a high-end luxury product, so their traditional merchandising method focused on showing the most attractive, highest priced items first. Does this merchandising work well on their eCommerce site?
To try and evaluate the viability of the current price-point based merchandising, we evaluated the sorting methods available on the site, engagement of product positions on category & search result pages, pageviews for products, add-to-cart rates for products, etc. Analysis showed that the products placed highest in the default sort were not performing as well as lower placed products. We theorized that although successful in their more traditional spaces, price alone is not the primary deciding factor for web shoppers.
Screenshot of a typical product grid:
We partnered with an outside firm that provides merchandising optimization technology. This technology builds multiple shopper profiles based on how frequently shoppers view a product, how long shoppers interact with a product, and how often a product is added to the cart. Based on those profiles, a sorting algorithm that weighted products dynamically by accounting for both user activity and manual merchandising was developed. This would allow the retailer to promote products that showed more user interest while balancing their own internal merchandising concerns, ie. price point, stock seasonal demand, etc. The multi-faceted sort algorithm should result in higher engagement and flow into product pages.
The experiment was run as a 50/50 split for 2 weeks. Although 2 weeks sounds like a short duration for an experiment, this particular client has very heavily weighted seasonal traffic, with approximately 85% of their yearly site traffic seen during that season. Approximately 5% of the site's yearly traffic was placed into the experiment over that 2 week period.
The multi-faceted sort resulted in the following gains on key metrics:
- Product Link Engagement - 34.51% conversion rate (+6.5%)
- Order Conversion - 1.8% conversion rate (+16.7%)
- Revenue - $4.38 revenue per visitor (+7.5%)
Product link engagement tracked clicks on links to product pages from within the product grid. This was used to gauge overall effectiveness of the presentation order in driving traffic to products. Order Conversion and Revenue helped to determine user behavior beyond initial interest. After all, getting more eyeballs on the products is great, but ensuring follow-through to purchase is ideal.
Product sort orders such alphabetical product arrangement or placement of the most attractive or highest priced products first can be logical, but not optimal for how web shoppers behave. Traditional brick-and-mortar or catalog techniques may not be effective, either. There are products that could capture your shoppers' attention and time, but have a very low add-to-cart ratio. In other cases, certain merchandise can have a high add-to-cart but are rarely purchased (even with the best abandon cart strategies). Curated merchandising and testing helps shoppers find the products they actually want to purchase.
We are ecstatic for our client, as they look to gain nearly $1 million more per year from the new Merchandising Optimization benefits. Without testing, attributing the wins and quantifying the gains of a new marketing strategy, new site design, or new technology can be extremely difficult. Considering the investment in new technology and the effort your team puts into operating all of your point solutions, testing new solutions is absolutely critical to long-term vitality and success.
I'm curious, did you find that the seasonal traffic is reflective of the general traffic? In other words, were the changes that generated the lift during seasonal periods still beneficial after the non-seasonal rush?