Recommender Systems
A recommender system looks at your historical data — past bids, project outcomes, supplier performance, crew productivity — and surfaces patterns that help you make better decisions on the next project. It's not magic. It's pattern recognition applied to the data you already have but aren't fully using.
How It Works
The system ingests your historical project data, learns what "good" looks like across different dimensions (cost accuracy, win rate, supplier reliability), and recommends actions based on those patterns.
What I Build Recommender Systems For
1. Bid Strategy & Proposal Optimization
The system analyzes your past proposals — wins and losses — to identify what made the difference. Was it pricing? Scope approach? The team you proposed? Client relationship history? It then recommends adjustments for the current bid based on those patterns.
In practice:
- Before submitting a bid, the system compares your proposed pricing against comparable past projects and flags significant deviations
- It identifies which team members have the highest win rate with this type of client or project
- It surfaces scope items you commonly include on winning bids that might be missing from the current proposal
2. Supplier & Subcontractor Selection
Instead of relying on the same three subs you've always used, the system ranks your subcontractor pool based on actual performance data: schedule adherence, change order frequency, safety record, and price accuracy. It factors in the specific project type and conditions to recommend the best fits.
In practice:
- When you need a mechanical sub for a hospital project, the system shows you which subs have the best track record on similar institutional work — not just who's cheapest
- It flags subs whose pricing has been consistently off (high change order rates), even if their base bids look competitive
- It identifies subs you haven't used in a while who performed well on past projects — preventing the "out of sight, out of mind" problem