Adding a new Recommendation
Navigate to "Marketing/Marketing Automation/Product Recommendations" to manage your product recommendations.
Adding - Types of Recommendations
|Product With Common Attributes||Lists products viewed by the customer that share the most viewed layered attribute. If a customer sees a lot of Nokia smartphones, recommended products will include all Nokia products.|
|Related Products from Last Completed Order||Lists all products that are related to products in last order. If I bought a smartphone and that smartphone had related products, those products will be displayed|
|Related Products from Previous Completed Orders||The same as above, but for all previous orders|
|Products in the Shopping Cart||Lists products in abandoned cart|
| Products in a Category||Lists products from a specific category|
| Products in the Customer Wish List||Lists products in the subscriber wish list|
|Most Viewed Products||Selects customer’s most views products.|
|New Products||Lists most recently added products to your store. These are the products that are marked “Set as New From” in your catalog. This option is not per subscriber, but rather global. All subscribers will see the same products|
|Recommendation Engine ||Recommends products based on our recommendation engine. More information below|
In the example below we are adding with the option "Product With Common Attributes" selected, but apart from "Recommendation Engine", options between other types are very similar
When you select the option “Recommendation Engine”, a series of options will appear to let you fine-tuning settings for the recommendation engine.
Widget options for product recommendations explained
|How wide should we search?|
(Please read the section below:
How do we build these recommendations")
|Recommend Based on Customer|
|Use Segments?|| If you want the information to be restricted to segments from which the customer is part of.|
How do we build these recommendations
We build metadata tables from information collected from the previous orders. Let’s consider the product Z. We gather information about all previous orders were the product Z was bought, and we check for the most purchased products along with product Z. But we don’t collect this information only at a global level. We collect this information for each country, customer gender, customer age and customer region. It’s easy to understand that different types of customers have different behaviors.
Adding to the segments we mention, we also build product recommendations for segments with the appropriate option selected when you create them (use in reports).
If you have a segment with customer who made more than 10 orders in the last year, you can combine that segment with the results from the country segment. You now can present recommendations that take in account the number of order and location of the customer. This means a customer, for example Spain, who made more than 10 purchases last year from the same product has a French customer, will potentially see different recommendations.
Not only we built recommendations based on what other customers purchase in the same order, but recommendations based on what customer bought in the next order. An example. A customer buys an Xbox. When we build recommendations for the Xbox product, we are not going to take in consideration products that were in the shopping cart when the Xbox was bought, but we are going to find out the most purchased product after customers ordered ad Xbox. With this recommendation you can try to predict market tendencies. Product recommendations in this type of settings will display products that customers did not think they would need when they first bought the Xbox. We might be talking about another controller, a case, some cleaning products, etc.
But we have another built recommendation. It’s what’s called “second level”. This name comes from the fact the recommendations are not built from the most purchases' product in the same order, but from the most ordered product in the order the recommendation result. Sounds complicated, but basically, and considering the Xbox example, we are not recommending based on the product that was most purchased with the console, but we are going to find out what was the most purchased product with the Xbox and then build a recommendation for that product and not the Xbox. The goal of this “second level” recommendation is to return items that, even though might not be directly connected to the Xbox, is commonly enough bought between the same customer that is to be considered at least of some interest in the customer.
The extension comes with a default template for this widget, based on the “New Products” widget from Magento®.
You can add your templates and specify the path in the template field when adding the widget.
This extension uses Magento log to select the appropriate products for each customer for the options: Product Attributes and Product Views. If you have large tables with millions of records, page loading may be affected.
Adding Recommendations to your store pages / emails
Green Flying Panda provides a widget that can be used to display products to users, subscribed to the Newsletter or not, that is based on each customer activity.
While in the editor, press the "Widget" button and from the window choose the option "Panda Sales Automation - Product Recommendation".
And then choose the corresponding Recommendation Entry.