Accelerating Digital Transformation through Enterprise Data Lake Solution
Enabling high value extraction from insights for a top cosmetic client through automation and machine learning
About the customer
The customer is one of the Top 10 cosmetic companies in the world and engages in the manufacturing, marketing, sale, and distribution of branded beauty products.The company is focused on deploying Machine learning techniques to improve their cash flow prediction process. One such use case is of Building Enterprise wide Data Lake solution for automating the need of predicting future cash flows and making insights more consumable. They contacted Brillio to get benefits of our expertise in Machine learning and Data Visualization space to execute the task.
In the consumer products manufacturing business one has to deal with lots of suppliers, vendors and specialists for certain process. The situation demands accuracy in terms of predicting the future cash flow because it gives one the current cash position of the company. Devising an AI-based solution for predicting future cash flow would thus enable an automated and reliable way. The problem is ripe for machine learning disruption. The key challenge lies in designing a solution which is efficient, consistent, accurate and adaptable in predicting future cash flow.
Step 1: Data Ingestion
Ingesting data from SAP BW4HANA into Azure using MDX queries
Setting up complete environment, from Azure DLS, Databricks, SQL DB, VM, Logic Apps, DevOps, ADF following security and folder structures, as per guidelines
Used Azure Databricks for enabling Data Transformation and Machine Learning usage
Setup Monitoring and Logging across the ETL process to detect any failures
Transform hierarchical information of various dimensions into consumable format for users
Provided dynamic table updates using SFTP with instant auto refresh in PowerBI
Step 2: Storage, Compute & Analyze – Building Machine learning model
Prepared ML model which predicts future payments date
Provides cash flow forecast using Random Forest algorithm, with 79% accuracy.
Used Pyspark, Azure Databricks, MLib package, and pipeline feature to execute repeatable codes
Provided outputs to PowerBI in a consumable format to further enable the easy to digest visualizations on the output
Step 3: Making Data ready for consumption
Transformed basic dashboards into interactive valuable dashboards
Provided frontend homepage like SAP for ease of use to users
PowerBI refreshes is triggered by Azure
Added features such as bookmarks, toggle, date scrollers etc in the dashboard
Prepared dashboards which transform ML outputs in consumable visualization
Provided users with PowerBI App, and trained users on its
Business Benefits & Impact
Achieved 81% optimization
Transformed a completely manual effort process (8 hours) and refreshed it into a completely automatic process (1.5 hours)
Real time Insights
Auto-refresh helps in extraction of latest data for discussion with client customers. The dashboard is always updated.
79% Forecast model accuracy
Cash Forecasted numbers are now based on statistical insights rather than the judgement of the business user.
Effective PowerBI dashboards enabling better business decisions and augmented user experience