Week after week we couldn’t believe the results. In a true A/B test, our new behavioral merchandising system bested our status quo, generating 6-8% more sales for our online store. At first, we were stumped as to why; we thought we’d optimized everything about our site: We had recommendations and social images on our product pages, we personalized our site search, we used faceted navs to allow for fast filtering on category pages, and we spent untold hours analyzing heatmaps of our homepage.
But in all those efforts, we overlooked one key fact: 80% of our site’s traffic visited a category page (you know, the pages that list all the products within a collection); however, we spent no time figuring out what to show visitors first. Sometimes products were out of stock. Sometimes they weren’t relevant to what a shopper really wanted. It was like if Google just spit out random search results based on the latest additions to the internet. Not a recipe for success.
So when we started putting our best products at the top of our collection pages, our sales jumped. If the steps below sound a bit complicated, don’t worry - Entaice handles everything for you (it’s free and easy to install).
Step 1: Set up your tracking
First things first, if you’re going to use data to merchandize your site, you’re going to need to track it. You’ll need a tracking pixel that lets you capture event data - specifically, you’ll want to track how many times a product was viewed, where the view originated from (i.e. which page was the visitor on before the product detail page), how many orders resulted from those views, their total price, and sku-level inventory. If it’s available, it’s also helpful to have a userid.
Step 2: Put all your data into a database
As you collect data, you’ll want to store it in a database for retrieval and analysis. Today you have a bunch of options through Google Cloud Platform, Microsoft Azure, and AWS; but, back in the day, we stored everything in an on prem (!) warehouse through our 3rd party analytics provider.
Step 3: Rank all your products by collection by likelihood to buy
Now that we’ve got all the data in once place, it’s time to rank your catalog so the products with the highest probability to sell show up first. There are a bunch of ways to do this with data science, but if you’re just getting started here’s a shortcut that will give you a head start. Just use the product-level conversion rate (i.e. orders / number of product views). For products without a lot of views, add a random component to push them higher up the page to see if they’re successful. Rank them by collection.
Step 4: Apply Inventory Filters
Great, now we have all of our products ranked for every collection. Next up, we want to apply an inventory filter to make sure we aren’t showing products that are out of stock. Depending on how often you’re updating your rankings, calculate the amount of inventory you’ll need in key sizes to last through the next update. For example, if you know a top selling sneaker sells 15 pairs of size 10 per day, and you’re updating your rankings daily, make sure there are at least 15 pairs in stock; otherwise set the product’s ranking to 99999 so it shows up at the end of a collection.
Step 5: Upload your new product sort to your ecommerce CMS
Grab the results of your ranking algorithm and inventory filter, and upload them to your store’s CMS on a schedule.