With the right e-commerce data science to success

THE STEADY FURTHER DEVELOPEMENT OF OUR RECOMMENDATION ENGINE.

Our e-commerce data sciencefor relevant recommendations

Machine Learning, Data Mining, Big Data, Predictive Analytics …. not just slogans, but the craft of our e-commerce data science. Our data scientists are using their know-how to optimize our Recommendation Engine every day to generate relevant recommendations. Theories for the online shop and the e-mail marketing are developed, tested and implemented.

How the further development of our Recommendation Engine works

Analysis of the online shops

We analyze the daily happenings in the online shop by looking into the business areas and products of the customer.

Development of theories

We use our collected knowledge from the field of consumer psychology to develop new theories that lead to even better recommendations. Following this, we translate the theories into the design of our algorithm.

Fine-tuning the algorithm

A specific target is defined which is to be optimized within a roughly defined framework. This makes it possible to quickly determine which theory is the best and which parameter settings are used.

Verification of the theory

We verify the effectiveness of the obtained theory with regard to the status quo via A/B tests. Here, we conduct statistical measures to achieve a high level of decision-making accuracy.

Data Mining and Machine Learningimportant is the interaction

Data Mining

_________

Data mining is about explorative data analysis. We use statistical methods to gain valuable insights from the data (big data).

Machine Learning

_________

In the field of machine learning, artificial intelligence is at home. Methods and techniques from Supervised and Reinforcement Learning are used to make our Recommendation Engine self-learning.

Clasical statistics

_________

This is a quantitative data evaluation. It is analyzed the occurrence of a quantity, under certain conditions, such as, the average amount of sales. Thus, we check our theories within the A/B tests for randomness or pattern.

Supervised Learning

_________

An input variable and a target variable are defined. The algorithm must rise from the input variable to the target size. To this end, he is given examples to learn how to achieve this goal.

Bayessche statistic

_________

Here, the probability is calculated between two elements, for example, the probability that it will rain when a black cloud occurs (computation of compound probabilities). In other words, when someone buys product A, how likely is it that he is buying product B? Thus, we check our theories within the A/B tests for randomness or pattern.

Reinforcement Learning

_________

Learning by doing with subsequent adjustment, based on the result. This is how the algorithm learns what works best.

Output

The right combination of techniques and methods in the field of data mining and machine learning, combined with a high degree of consumer psychology, enable us to make valuable predictions for the future (Predictive Analytics). As a result, our e-commerce data science team is continually developing our recommendation engine.

Where do all the data come from?

The data are obtained in real time from the click and purchase behavior in the online shop and the e-mail. However, further data can also be incorporated, for example from:

a CRM system
the stationary trade (purchase data, customer cards, etc.)
the product attributes in the product catalog
various external data sources (e.g. return data from a merchandise management system)

Our data science teamThe experts in the background

The Data Science team is closely connected to the Customer Care team. Even if you do not speak directly to the Data Science team, they follow the tickets and help with advice and action. The expertise flows into general settings and the customizing.