Customer 360

By for June 23, 2017

Report Abuse
A deep understanding between customer interests and purchasing patterns is a critical component of any retail business intelligence operation. This solution implements a process of aggregating customer data into a “360 degree” profile, and uses advanced machine learning models backed by the reliability and processing power of Azure to provide predictive insights on simulated customers.
> **Note:** If you have already deployed this solution, click [here]( to view your deployment. > **Estimated Daily Cost:** [$369.25]( > For more details on how this solution is built, visit the solution guide in [GitHub][1]. ### Estimated Provisioning Time: 20 Minutes A typical retail business collects customer data through a variety of channels, including web-browsing patterns, purchase behaviors, demographics, and other session-based web data. Some of the data originates from core business operations, but other data must be pulled and joined from external sources like partners, manufacturers, public domain, etc. Many businesses leverage only a small portion of the available data, but in order to maximize ROI, a business must integrate relevant data from all sources. Traditionally, the integration of external, heterogeneous data sources into a shared data processing engine has required significant effort and resources to setup. This solution describes a simple, scalable approach to integrating analytics and machine learning to predict customer purchasing activity. The Customer 360 Profile solution addresses the above problems by: * Uniformly accessing data from multiple data sources while minimizing data movement and system complexity in order to boost performance. * Performing ETL and feature engineering needed to use a predictive Machine Learning model. * Creating a comprehensive customer 360 profile enriched by predictive analytics running across a distributed system backed by Microsoft R Server and Azure HDInsight. ## Solution Architecture ![Architecture diagram][IMG1] 1. A `Data Generator` pipes simulated customer events to an `Event Hub` 1. A `Stream Analytics` job reads from the EventHub, performs aggregations, and persists time-grouped data to an `Azure Storage Blob` 1. A `Spark` job running in `HDInsight` merges the latest customer browsing data with historical purchase and demographic data to build a consolidated user profile 1. A second `Spark` job scores each customer profile against a machine learning model to predict future purchasing patterns (i.e., is a given customer likely to make a purchase in the next 30 days, and if so, in which product category?). 1. Predictions and other profile data are visualized and shared as charts and tables in Power BI Online. [IMG1]: [1]: ## Disclaimer ©2017 Microsoft Corporation. All rights reserved. This information is provided "as-is" and may change without notice. Microsoft makes no warranties, express or implied, with respect to the information provided here. Third party data was used to generate the Solution. You are responsible for respecting the rights of others, including procuring and complying with relevant licenses in order to create similar datasets. ![ ](