Problem Statement

​

Our client, a prominent name in commercial construction, recognized that their pricing model wasn’t capturing enough value. They noticed lost opportunities, where bids were either too high to win or too low to maintain healthy margins. The solution had to address this delicate balance and empower estimators to make strategic pricing decisions that aligned with the company’s profitability goals.

Beyond that, they needed a tool simple enough for non-technical users to operate yet sophisticated enough to deliver accurate, data-backed guidance. They sought a system that could adapt to project, customer, and market variations—one that would make every bid a calculated move towards both competitiveness and profitability.

​

This is where Phenx Machine Learning Technologies came in, building a data-driven pricing engine that turned historical data into strategic pricing recommendations. The tool’s goal: empower estimators with a self-reliant pricing strategy that balances competitiveness with profitability, putting actionable insights into their hands and making each bid a more predictable step toward growth.

​​

Project Overview

​

We began by analyzing years of the client’s project data to uncover trends and correlations that could drive more effective pricing decisions. Our approach wasn’t about creating a generic pricing model—it was about designing one that aligned with the specific nuances of each project type, customer profile, and location.

​

The foundation of our solution was a custom LightGBM model, designed to handle the unique challenges of project-based pricing in commercial construction. Segmenting the data by key factors (30+ variables) such as job type, project size, geographic location, and macroeconomic variables, we highlighted how each segment’s win rates and profit margins responded to pricing changes.

​

Data alone, however, was not enough. We applied a Bayesian approach to estimate win probabilities with confidence, even in data-scarce areas. This allowed the model to generate reliable recommendations, enhancing the client’s pricing precision and giving them a competitive edge in unpredictable conditions.

​

The result was an intuitive tool that estimators could use without requiring data science expertise. With a few inputs, they could see optimized pricing adjustments backed by insights into win probability and profit impact. Seamlessly fitting into their existing workflow, this tool transformed the way they approached pricing—making it a valuable asset for decision-makers.

​​

Solution Detail

​

To meet the client’s unique needs, we crafted a pricing engine with distinct components that adapted to the complexities of their industry:

• Clustering and Segmentation: We started by clustering projects based on key similarities such as job type, location, and project size. This structured approach reduced pricing variability while allowing us to tailor pricing adjustments to specific project categories.

​

• Population Control for Reliability: We set minimum sample sizes per cluster, ensuring each segment captured unique project attributes without sacrificing data reliability. This way, even sparse data was used meaningfully without compromising the model’s precision.

​

• Bayesian Price Adjustment: By estimating win probabilities with a Bayesian model, we managed data uncertainty. This approach allowed estimators to make pricing decisions confidently, backed by rigorous statistical support, even for projects with limited historical data.

​

• Intuitive, User-Friendly Interface: Designed with non-technical users in mind, the tool provided actionable pricing recommendations. Clear insights into win rates and profit impact enabled estimators to make quick adjustments and see real-time guidance. For executives, this meant a reduced need for technical support and empowered decision-making within the team.

​

​