Switches

Switch and User Segment

Switch and User Segment Node has the same parameters, the only difference between them that Switch has multiple inputs but executes only one of them, User Segment has one input and multiple outputs, but activates only one of its outputs. The same logic can be implemented with both blocks:

The Predicates used to select the subgraph which is activated is the same for both Nodes.

Each block can have a name and also a Predicate. The Predicate decides if a block is selected or not. Always only one block is selected. If more than one Predicate is true for a certain situation, than the first from the top is activated. If no Predicate is true, then another block is activated. Predicates can be specified based on user property, context parameter or user events. If a user event is selected, the events can be accurately specified, e.g by filtering events happening only in the last 24 hours in a certain item category.

Selecting “User Events” allows you to select users based on their past behavior or history. It enables you to select different event types like: Buy, View, Add to Cart, or All Available Events. For example, selecting “Buy” will display all of the people who ever bought something on a certain site. Selecting all available event types will display all the people who have previously visited the site.

There are four toggle buttons called “Item”, "Timing", “Event”, and "Number limit". Clicking on the “Item” button allows you to create a more detailed filter. Using the drop-down menus, one can narrow down the range of site visitors. For example, select price below “Item” and then “greater” below “Match type”. Now type in a value below “Item Property Value” to complete the equation. This means that you are filtering for everyone who has ever bought an item of $1,000 or more. Click on the “Add” button to make this filter become active.

Under "Timing" you are able to filter for a certain time period, for example, users, who bought something in the last two days.

Clicking on the “Event” box enables you to define a period of time during which certain actions occur. For example, you can select the users who bought an item during the past five days. It is also possible to select the users who bought something more than a month ago.

Under "Number limit" you can keep on narrowing down your target group by setting a minimum (and maximum) number of events from the specified kind - for instance, users, who opened 3 or more product pages in the last 10 minutes.

The property “Log Decision Name” is used for debugging: if it’s defined, the name of the selected block is saved in the reco engine logs with the specified name for further analyses.

A/B Test

Yusp recommendation engine supports A/B testing. A/B testing is crucial for optimizing the recommendation logic. It enables to run different recommendation logic for different user groups at the same time in the same scenario, and to measure the performance of each logic. If an A/B Test node is used, it automatically generates the measurements for KPIs specified in the BRD. The A/B test measurements are only available after the A/B test was published in a touchpoint.

In the A/B test node, any number of groups can be specified. The customers are randomly divided between the groups according the size of the ranges specified with the sliders next to each group. A/B Test always selects one of its inputs so the different versions to be tested has to be placed before the A/B Test node in the data flow graph.