ClientAs a large retailer our client wanted to understand customer behavior and thus needed to modernize its cloud infrastructure to develop a Customer 360 platform. The increasing variety of siloed data source made it impossible for the company to manage efficiently.They were aware of the need of developing a smooth data pipeline with well-engineered services and triggers to establish a robust software as a service architecture. The latter was required to create a unified and centralized customer database that includes all touch-points and interactions with their product. The existing data management approach had no centralized database with siloed data stemming from multiple sources in diverse formats.For that, the existing data ecosystem had to be reorganized into a consolidated cloud SaaS architecture to centralize customer data from all sources and make it available for further analysis.
Challenge: integrate heterogeneous data sources to obtain a unified 360-degree view of a customerThe main bottleneck on the way to a unified data flow was the unstructured and siloed nature of data. The customer data resided in multiple sources, including SAP, Salesforce, API, Flat Files, CSV, and others. Before building the SaaS product architecture, our team needed to transform the existing information, including related tasks of data cleaning, mapping, and merging.Further, a customer data platform was required to be seamlessly integrated with each of the SaaS products. A CDP also needed advanced automation capabilities to build a coherent understanding of each customer.
Solution: custom CDP system based on MDM process and ML-based data matchingThe development process started with our team of Big data engineers analyzing the client requirements and subtleties of the current data architecture. The project was delivered in several phases and covered design, implementation, and deployment of SaaS application architecture.The following project tasks were performed to establish a comprehensive customer database for analysis:
- Different data types and formats was pulled into common data format with Azure Data Factory
- The storage system was established within Azure Data lake
- Data was loaded into ADLS Gen2 Azure Blob as a raw layer in CDM format
- Cleaning, mapping, and merging was performed with Azure DataBricks
- Profiling was done to match the records from different entities and create a unified profile
- Map match and merge was performed for three different pipelines for a three rule-based solution based on
- Machine Learning algorithms coupled with Graph Algorithms.
Result: a custom designed SaaS data architecture to ensure a proactive and actionable customer strategyThe Essid Solutions team created a CDP system to get the data from multiple sources and integrate it with other SaaS business applications. We combined the data using the MDM process and Machine Learning matching to show the complete 360-degree view of the customer.The final merged data can then be presented via an appropriate UI – effectively a simple dashboard, which displays a unique granular customer profile in terms of many significant KPIs and analytics models.As a result, the client can now aggregate and organize customer data across a variety of touchpoints and use this data for further analysis, targeted marketing efforts, and other personalized services.