Company Type
McClarin Plastics, a custom plastics manufacturer with three plant locations; CPQ needed for growing custom-product complexity — the client believed no commonality existed across historical jobs.
What Drove the Need
A CPQ solution required standard production profiles; the client's null hypothesis was that every job was unique
Excel was insufficient for the combinatorics
Business Impact
Null hypothesis rejected — 5 core profiles encompass all job types; at one location six routing descriptions accounted for 91.3% of recorded hours
Future-state QTC and PLM approved by C-level and departmental leadership
Systems/Data Integration
Read-only extraction from the S2K ERP; results feed the proposed CPQ design and future-state Quote-to-Cash and PLM processes
Design
Method escalated deliberately: S2K tables (item master, customers, manufacturing orders, routings) exported, decoded via the data dictionary, staged in Access, converted to SQL Server 2008
SSIS stage: Term Extraction (noun and noun-phrase, frequency-scored) over routing descriptions → Term Lookup re-tagging every routing record → Data Profiling Task across all columns to expose distributions, outliers, and relationships
SSAS stage: mining structures over manufacturing orders and routings; iterative dataset reshaping per domain to isolate drivers; the Excel add-in used for fast passes to steer the heavier models
Labor hours chosen as the driver because hours and cost correlate ~1:1; hypothesis qualified against academic literature on data-mining-generated generic BOMs
Technology Stack
Microsoft SQL Server 2008; SQL Server Integration Services; SQL Server Analysis Services Data Mining
Excel Data Mining add-in; MS Access staging
VAI S2K ERP on IBM i; IBM WebSphere Portal
Job Shop Profiling Application (ML-Generated Production Profiles)
Machine-learning generation of production profiles (type, cost, cycle time) from S2K production history: SSIS term extraction and profiling, then SSAS clustering, neural network, Naive Bayes, decision-tree, and basket-analysis models steered by the Excel Data Mining add-in.
