Provide individual product recommendations based on the customer's profile and the relevant context. That way, the user always gets the best recommendation at the right time at the right place.

The Basis: A Recommendation Engine

The Basis: A Recommendation Engine

Based on click and conversion rates, the self-learning algorithm continuously optimizes the display of products in the background. Previously created layout variants (e.g. InPage banner versus overlay versus exit intent) as well as various product logics can be tested and different algorithms can be saved.

Product Logics for Appropriate Recommendations

In order to be able to display recommendations precisely, product logics must be defined in the Recommendation Engine. For example, the "Recommended Products" logic selects products according to a sophisticated algorithm that responds to the user's surfing behavior on the basis of his statistical twins and also includes sales of the respective products.

Product Logics for Appropriate Recommendations
Recommendations Based on Similarity

Recommendations Based on Similarity

Online shoppers usually look at several products of the same category and compare them with each other. Recommendations can be used to facilitate this search by suggesting and displaying similar articles as recommendations. However, recommendations based on product similarities are very static and ,therefore, not useful for every offer.

"Customers also bought…"-Recommendations

A further possibility to inspire users are "Customers also bought... "-recommendations. In this case, the shop visitor is shown further articles ordered by other customers in connection with the product they were looking for. This type of product recommendation is particularly useful for shops with a homogeneous product portfolio.

Manually Assigned Recommendations

Manually Assigned Recommendations

Slightly more (time) consuming are product recommendations by means of manual allocation in order to motivate customers to buy impulsively through individual offers. Based on human perception, it is decided what really fits together or what could complement the product in question. In this way, potential customers can be offered accessories or bundle products, for example.

Recommendations Based on Statistical Twins

Recommendations relating to the customer journey are particularly individual. An intelligent algorithm decides on the appropriate product recommendation. Buyers are clustered into groups and their purchasing behavior and shopping basket are analyzed. Based on this data, customer-specific suggestions are developed.

Recommendations Based on Statistical Twins

Recommendations in Use

How mydays increased the conversion rate by 6.9 % using the trbo Recommendation Engine

mydays products vary strongly from region to region. In order to address both new and returning users, mydays decided to test the trbo-Recommendation Engine. Two customer segments were formed and provided with different playout logics. The result: the trbo-recommendations improved the conversion rate by 6.9 percent.


Increase in Conversion Rate

Recommendations: The technical basis

Dynamic Segmentation
Geo- & Weather-Data
Machine Learning
Click-In Channels
Big Data

The technical basis for suitable recommendations is a clean product feed that keeps all products in stock as up-to-date as possible. In addition, a clean product and conversion tracking must be guaranteed. Based on the selected playout, user characteristics (e.g. traffic channel or weather) flow into the logic, which is processed by an intelligent algorithm.


Which recommendations receives the best echo from which user? To answer this question, you should use (dynamic) A/B- & multivariate tests. They enable exact statements as to whether a certain recommendation logic is useful or not. Learn more about A/B-testing now

trbo GmbH
Leopoldstr. 41 | 80802 Munich, Germany
+49 (0) 89 215 38 68 0 |

trbo Inc.
429 Lenox Ave | Miami Beach, FL 33139 United States
1-844-200-7222 (toll free) |

Request demo