You don't have to look past the home page of Amazon.com to find literally dozens of algorithms, all hard at work personalizing results for you. Behind the scenes Amazon has nearly 1000 data scientists working across their sites and those experts manage uncounted thousands of algorithms.
No one else in e-Commerce has the resources that Amazon has for data science talent, though some companies like Stitch-Fix, Etsy, and Jet.com are already investing heavily in their algorithms.
So where does that leave the bulk of e-Commerce? For years, most companies have made due with a few dynamic landing pages and a lot of wishful thinking. At times, the "place a pixel" black-box algorithm vendors were seen as the solution. But we now know they are just part of the problem for e-commerce leaders.
The good news is that we now have emerging Algorithm Management Systems (AMS) like Terrain that bring ease of use, transparency and control to the algorithm process. With Terrain, one motivated technical marketer can now play the role of an entire algorithms team. The 25 starter algorithms that follow can all be implemented in the first month of using an AMS like Terrain, so lets dive in!
The First Algorithm Every Site Needs
What do you display when you know nothing about your visitor? You need to deploy a Cold Start Algorithm.
A Cold Start Algorithm is designed to provide optimized product offers when you have negligible information about the user you are promoting to. It provides recommendations for Incognito visitors, new visitors, or people who are not logged in.
Rather than trying to predict what a cold start visitor wants to see, this algorithm optimizes results based on what you know about your products using factors like price, conversion rate, inventory and popularity.
AirBnB shows Experiences first before Homes for unknown site visitors
Segment algorithms promote products clustered around a user behavior. Some examples include:
- Event Driven - Onsite holiday promotion and marketing campaigns
- Preference based - Sending emails featuring products that match the email frequency preference
- Psychographic - Marketing clustered around a known interest, like having children.
Amazon knows me too well :-(
Lifecycle algorithms adjust content depending on where the customer is in their decision process. Examples include:
- First Time User - Recommendations aimed at a user who has never made a purchase.
- Sign up email - Special offers to get a new registrant to engage in a behavior
- Loyal User - Informed recommendations based on purchase history
- Churned User - Last ditch offer to bring dormant users back to loyalty.
- Abandoned Cart - Email marketing that is aware that a user almost purchased items.
Levi's knows I want those pants
Promotional algorithms optimize results for customers who anchor their purchases on discounts and promotions. Examples include:
- Discount Buyer - Marketing high margin products to a customer who only buys with coupons and discounts
- Deal Offers - Customized recommendations tied to specific promotions and deals
- Closeouts - Recommendations aimed at clearing remnant inventory
Nike brings you back to their app
Trending Algorithms promote products that are popular over a specified time period or other factor. Examples include:
- Hot Now - Offers focused on fresh and popular products and services
- New - Used to highlight products or services new to your catalog
- Top Sellers - Promoting the historically most popular products
- Near You - Offer popular products or other relevant categories that are close to the shopper
- Almost Gone - Inventory aware marketing intended to drive urgency
Netflix is a master at trending algorithms
Related Algorithms are algorithms that feature products you might like based on some previous behavior. Examples include:
- Because you purchased - Marketing for loyal customers with a purchase history.
- Similar to this - Recommendations aimed at getting users to make one more click.
- Cross-sell - Recommendations aware of products or categories that are frequently purchased together.
Zappos knows what I might like with those pants
Personalized algorithms feature products that are recommended for an individual user based on that user's data set. Examples include:
- For You - Highly targeted "just for you" marketing offers
- Close to you - Recommendations where location and proximity matter
- Social Proximity - Recommendations where your friend-network preferences matter
- Categories you might like - Used when you are trying to learn more about the customer.
Medium knows just the kind of article I want to read
At Terrain we believe that every company should have hundreds of algorithms. But regardless of where you end up, you can start with picking a few simple ideas and get them into production. We can help.