Our client is a leading distributor of specialist building products in Europe, with strong positions in its three core product areas of insulation and interiors, roofing and exteriors and air handling.
The company plays a critical role in the construction supply chain, ensuring that customers receive the right product, in the right place, at the right time. It operates from trading sites across the UK and Ireland and Mainland Europe, and employs around 9000 people. Its main countries of operation are the UK, France, and Germany.
Business Challenge
Our client is a multi-national group that has grown through acquisition and has been managed historically as a portfolio of businesses. Due to the nature of its operating model, business processes and systems were not always standardized and there was minimal integration of processes and data. For example, they had a multitude of standalone ERP systems across multiple geographies.
One of the crucial issues our client faced was the unique pricing approach because of the amount of transactions and market share they have. There were different CPQ (Configure, Price, Quote) processes supported by multiple systems across the company. Determining the best price for a particular deal or customer was challenging and there was no price control mechanism in place. Our client’s employees often would use gut instincts and experience when quoting a price to a customer, which resulted in products being sold for low or negative margins, impacting overall profit.
Specific goals our client had in working with SoftServe included:
- Leverage machine learning algorithms to provide flexible price recommendations to maximize margin dollars without the risk of losing business
- Analyze external pricing influences: market, demand, seasons
- Determine pricing structures for initial pricing and discount pricing, including rebates and historical deals
- Optimizing demand prediction for each category/SKU
Project Description
SoftServe analyzed our client’s business challenges and developed a Smart Price Calculator to calculate the ’right’ and competitive quote for every particular customer, whether B2B or B2C. The core of the calculator is a machine learning engine that calculates best deal price based on historical data, products, and customer characteristics.
During a four week Discovery PoC, an onsite was conducted to elicit the scope of requirements and PoC acceptance criteria, analyze as–is state from business and technical perspectives, and gather and investigate data sources. Subsequently, a PoC implementation was conducted on two main tracks: algorithms + R/Shiny application and UI/UX.
Value Delivered
SoftServe successfully delivered the following POC results and our client used these as the basis for further MVP pricing product development.
POC Results:
- Data understanding and aggregation, units of measure unification
- Machine learning-driven algorithms to prove the concept of pricing in the app back-end
- Pricing on the customer, branch, product and quantity level based on XGBoost model
- Demand prediction on the product category level
- Price break-down visualization, including simulated competitor price adjustments and rebates
- Customer and product historical marginality view
- R/Shiny dockerized application to assess the pricing engine, git repo
- Summary of the approach, data quality issues, MVP suggestions
Technology Stack
- R/Shiny, Python/iPython
- XGBoost
- Docker
- MS: Azure, SQL Server , SSIS
