You’re shopping online for a new rug for your living room, and you find the perfect one. You add it to your cart, and on the next screen you see product recommendations pop up for rug pads. Oh yeah! You do need a new one for your new rug, so you add that to your cart too. You’re a happy shopper and you got everything you needed without any pushy ads, pop ups, or sifting through a bunch of products you don’t want. Now, as the person responsible for website conversions on your own site, you want to provide that same experience for your own customers. You have your product recommendations tool, and here you reach the pivotal question: Am I thinking of “Complementary” product recommendations or “People Also Like” product recommendations? If these two deceptively similar algorithm types cause some confusion, you’re not alone. Here are the differences, and when you should use each.
Understanding Product Signatures
When we talk about product recommendation algorithms, we first need to understand what we mean by “product signature.” The signature is made up of all the amalgamated actions users take that lead to purchasing the product. For example, if before you bought your rug, you:
- Did a site search
- Looked at three other similar rugs
- Viewed the product details of the rug you ultimately bought
- Clicked on several product images
- Clicked on the product dimensions
- Added the item to your cart
These actions all add to the signature of that item. Now imagine that on a massive scale, where all the actions that all the customers who ever purchased that rug took come together. These patterns make up the product signature.
“People Also Like” Product Recommendations
Algorithms based on items other shoppers liked run based on product signature. In our scenario, a “People Also Like” algorithm would take the product signature of the run and compare that to the signatures of other items. It would then recommend items with the most similar signatures. It’s important to understand that this algorithm does not factor in each user’s shopping cart as a whole, it works by comparing individual items.
Complementary Product Recommendations
“Complementary” product recommendations are not based on product signature. Instead, the algorithm uses frequent pattern matching to compare shoppers’ complete carts. For example, when you select your rug, the algorithm compares the complete carts of every other instance that someone has purchased that rug. It does this on a massive scale, and returns items that are usually purchased with each other.
To learn more about SiteSpect, visit our website.
About Sally Hall
Sally Hall is an Optimization Consultant at SiteSpect, guiding SiteSpect users on the road to optimization. She has more than 10 years of experience as a web optimizer and testing manager for enterprise brands. She is based in Austin, Texas.