e-Marketer recommendation engine includes over 50 algorithms for use across websites, emails, and apps, ranging from simple but effective algorithms like “View It Again” to highly complex machine learning recommendations. All recommendations are responsive and can be displayed as popups, embedded into pages, emails or apps, with any HTML/CSS design you like.
Cross-selling Algorithms
Those who bought this also bought…
A go-to cross-selling algorithm commonly used on product pages. This recommendation shows items that were brought together with a particular product, whether it be the current product the visitor is viewing, or the product now in the cart, the last product added to the cart, recent purchase, all purchases, and more.
Those who viewed this also bought…
Another great cross-selling algorithm. This shows items that were viewed by visitors who viewed the current time during the same session, or by the same user at a later time. It can be based on the current product the visitor is viewing, the product now in the cart, the last product added to the cart, or a product category that the visitor is viewing (a mix of products in that category).
Those who bought from this category also bought…
This algorithm can be used for cross-sales on a product page, or also on product category pages, and shows the items most frequently bought from the category the visitor is currently viewing.
Cross-selling algorithm enhancements…
You can use the e-Marketer recommended option for this algorithm, which is to choose and recommend only items that were bought on the same session or order ID, or bought by the same user. You can also use e-Marketer ‘s machine learning cross-selling option, that factors in the likelihood of a user to buy a product, and many other variables besides quantities.
You have the option to adjust the algorithm with various limitations, such as to not recommend items unless there were 3 transactions where the items were bought together.
To further enhance, you can filter based on the price of the items; usually, you will want to show only items that are 20% less than the current product, to increase the likelihood of showing the accessories of the given product. You can also explicitly tag accessories, in your product feed.
Extra bonus: you may import transaction data from your offline channels to have more data to recommend and to jump-start your recommendations when you first add them to the site.
Best Sellers and Most Popular Algorithms
This category of algorithms shows the most popular or best-selling items from your site overall, or from various categories, and can be adjusted to different scenarios. They are commonly used on home pages and category pages, but may also proven useful in many other places.
Based on interest or category, you can present a mix of best-sellers only from categories/interests the visitor viewed, added to cart, or purchased.
Based on page type, you may show best-sellers from the currently viewed product category/interest and on category pages, show the best sellers from this category.
Timing is important: for a given time range, you may choose to show best sellers, most popular for all time, for the last X number of days, or you can use the “recently” option, which will first show the newest bestsellers from today, then from last 2 days, then the last 7days, etc. until all the requested items to show are filled.
New in stock
This algorithm displays items that are new in stock since the last visit, with the option to fine-tune for items added to the site in the last 7 days, before the last visit. You can also show only items that are based on visitor’s interests, or only from categories they have viewed, added to cart, or purchased.
Recommended for you
This machine learning based algorithm uses the totality of user data to find patterns in how users of various types interacted with all of the items in your catalog, and creates recommendations based on these insights. This can be an algorithm which creates highly relevant recommendations which a human creator would never know to create, but also requires a large quantity of data to make meaningful insights. For high-traffic sites, it can be a very effective algorithm, and is commonly used on home pages.
Up-sells recommended for you
Similarly machine-learning based, this alternative will emphasize items that are upsells for the same product, which should be explicitly labeled in your product feed/catalog.
Recent views/your most viewed items
This non-algorithmic recommendation simply shows the visitors most recently or frequently viewed items, and although simple, can result in a lot of sales by reminding visitors of items they want, especially when used on the homepage and landing pages, for when they return to the site, or in remarketing emails.
There are also many other recommendation algorithms, around 25 total intended for products, and another 25 for content.
For each of the recommendation algorithms, you can add filters such as:
Show only items that the price is greater or lower than X%, or a fixed amount lower, or show only products that are cheaper than the current product on product pages, or mixed items from different categories that the visitor interacted with, based on price, on the home page, for instance.
Based on real-time visitor location
If your product feed contains item locations, you can add a filter that recommends only items that are available within X miles from the visitor’s location. Location is based on IP address by default (accurate to the city level), but you can activate a e-Marketer action to the more accurate use browser location via Google map data, although this requires the visitor’s consent.
Location can also be factored into e-Marketer ’s machine learning algorithm to show items that are popular in certain locations, such as countries or regions, to customers who are visiting the site from that region.
Discount
e-Marketer allows you to show recommendations that highlight discounted items. You can filter to show only items that have X% discount or more, for instance.
Stock Availability
Show items only if available stock is more or less than X units.
Personal User Data
If your product contains age range or gender tags, you can filter based on either, assuming you set e-Marketer to have the visitor’s gender and/or age.
Any Current Product Attributes
You can filter based on any attribute you have in your product feed, such as interest, color, brand, or other tags.
For instance, you can show items that other visitors viewed, that have the same color or size as the current product the visitor is viewing.
You can also trigger or limit each of the recommendations to show only:
On the First Page in Session
Show the most viewed items based on previous behaviors, only on the first page of a return session.
Show Only to B2B/B2C Customers
If you have separate segments like B2B vs. B2C, you can limit the recommendations so that each Segment sees the products that are optimal for them.
Immediately After the Event
Show cross-sale items in a popup, for instance, right after the visitor clicks to add to cart.
Show On Exit
Suggest sending an email to the visitor with their viewed items, only when they try to exit, using an exit popup.
On Scroll to Bottom
Show recommendations only if the visitor scrolls to the bottom of a given page.
Social Proof and Feedback for Recommendations:
“Great choice, this product was bought 100 times last week!”
“This is low in stock, hurry up!”
“This is currently on sale for a 70% discount!”
Dynamic Reviews
Show product reviews dynamically from the visitor’s location, or based on categories they viewed. For instance, show visitors product reviews from those who are from nearby locations to their own.
