Procurement organizations today are drowning in data but starving for insight. They have dashboards that show what happened last quarter, but very few have decision engines that recommend what to do next.
This project bridges that gap.
I built a decision intelligence platform that transforms raw ERP data—vendors, invoices, purchase orders—into actionable procurement recommendations. The platform combines SQL scoring models with Tableau executive dashboards to help procurement leaders answer three critical questions:
Which suppliers are creating operational risk?
Which products should be under contract?
Which suppliers deserve more business?
The result is a unified platform that moves procurement from reactive reporting to proactive decision-making, with $4.0M+ in illustrative annual value opportunities.
This is like 3 different case studies integrated into one.
The Challenge
Procurement teams manage hundreds of suppliers and thousands of invoices, yet most analytics remain descriptive—reporting what happened rather than recommending what to do next. This creates several significant gaps.
First, decision-making is fragmented. Supplier risk, contract optimization, and supplier development are typically addressed in isolation, with no unified framework connecting them.
Second, static risk labels—"Low," "Medium," "High"—are assigned during supplier onboarding and rarely updated, failing to capture operational friction like invoice errors and payment rejections.
Third, procurement dashboards are descriptive rather than prescriptive. They show historical KPIs but do not recommend actions.
Fourth, concentration risk is often invisible—suppliers with limited product diversity create dependency risk that is rarely tracked.
Finally, historical spend analysis can be dangerously misleading, recommending contracts for products with collapsing demand.
The organization is a multinational company that had 88 suppliers in USD, 2,168 invoices, and $626.9M in annual spend—yet lacked visibility into operational risk, contract opportunities, and supplier portfolio health.
I set out to build a decision intelligence platform that transforms ERP data into actionable procurement recommendations. The platform needed to integrate three decision engines into a unified architecture, use transaction-level data rather than static labels, employ fixed business thresholds for stable and defensible scoring, and produce executive-ready outputs that support procurement decisions.
The goal was not just to build dashboards, but to create a decision engine that recommends specific actions to procurement leaders.
The platform was built using a modern analytics stack. SQL (MySQL) handled all data extraction, transformation, and scoring logic using Common Table Expressions and window functions. Tableau provided the visualization layer, delivering executive dashboards with KPI cards, scatter plots, treemaps, and Pareto charts. Microsoft Excel supported data exploration and validation during development. All documentation was written in Markdown and PDF, with Git and GitHub managing version control and portfolio sharing.
This engine answers: Which suppliers create operational risk?
It uses a weighted scorecard combining spend exposure (30%), invoice error rate (30%), payment rejection rate (20%), and inherent risk level (20%). Suppliers scoring 70 or above are flagged for replacement; those between 40 and 69 require monitoring; and those below 40 are safe to retain.
The engine revealed critical insights: the sole "High Risk" vendor actually performed flawlessly with zero errors and 3.6% rejections, while a "Low Risk" vendor drove the highest error rate at 10.71%, 4.7 times above average.
This engine answers: Which products should be under contract?
It evaluates products across five dimensions: spend impact (30%), purchase frequency (25%), demand predictability (20%), price volatility (15%), and supplier concentration (10%). A Phase-Out Override prevents contracting products with collapsed demand—those with YoY decline exceeding 50% and 2025 spend below 25% of 2023 spend.
The engine identified one immediate contract candidate—Power Tools—with $2.47M in spend, 130% YoY growth, and $198K in estimated annual savings. It also flagged 35+ products for phase-out, including one with a 93.8% demand decline, preventing what would have been a costly contracting error.
This engine answers: Which suppliers deserve more business?
It uses a 5-dimension scoring model: Business Importance (20%), Operational Quality (30%), Internal Processing Efficiency (20%), Strategic Position (20%), and Risk (10%). Strategic intelligence flags—concentration risk, consolidation opportunity, performance declining—overlay the score to drive specific actions.
The engine revealed that 10 out of 12 suppliers have limited product diversity (two categories or fewer), exposing $87.7M to concentration risk. Three suppliers were identified as preferred partners with perfect accuracy and strong performance.
The Tableau dashboards deliver the insights at a glance. The Supplier Risk Matrix combines spend, risk score, and error rate in a scatter plot with clear decision thresholds. The Spend Exposure Treemap visualizes where the $627M is concentrated. The Performance vs. Benchmarks chart compares suppliers to operational averages. The Contract Priority Scatter identifies products for immediate negotiation. The Savings Concentration Pareto shows how many products drive 80% of savings. The Supplier Portfolio Quadrant segments suppliers by importance and quality. And the KPI Cards provide the headline numbers executives need in seconds.
The project unfolded in three phases.
First came data exploration and SQL modeling. I extracted and cleaned ERP data from vendor, invoice, and purchase order tables. I built three CTE-based scoring models—one for each decision engine—using weighted scoring with fixed business thresholds rather than dataset-relative normalization. I implemented guardrails like the Phase-Out Override and Low-Base Growth signals to prevent errors. The result was clean, aggregated data ready for visualization.
Second came dashboard development. I connected Tableau to the aggregated data and designed executive KPI cards to establish context. I built seven interactive dashboards, each answering a specific decision question. I added reference lines, annotations, and executive insights to make the visuals decision-ready.
Third came documentation and packaging. I wrote a three-page Executive Summary for leadership, documented the full methodology in a 40+ page technical report, created a GitHub repository with all SQL scripts and documentation, and prepared dashboard screenshots for the portfolio.
The platform analyzed 88 suppliers, 2,168 invoices, and $626.9M in annual spend. It identified 18 vendors requiring active monitoring, representing $174.8M in spend. It found one immediate contract candidate—Power Tools—with $198K in estimated annual savings. It flagged 10 suppliers with concentration risk, exposing $87.7M to single-product dependency. Across all three engines, the platform identified $4.0M+ in illustrative annual value opportunities.
The analysis also delivered three critical insights. First, static risk labels are unreliable: the sole "High Risk" vendor performed flawlessly, while a "Low Risk" vendor drove the highest error rate. Second, demand trends prevent poor decisions: the Phase-Out Override prevented contracting for 35+ dying products. Third, concentration risk dominates: 10 out of 12 suppliers have limited product diversity, exposing $87.7M to single-product dependency.
The project demonstrated advanced SQL skills including Common Table Expressions, window functions, complex aggregations, and NULLIF protection.
Tableau dashboard design skills included scatter plots, treemaps, Pareto charts, KPI cards, and diverging bar charts. Data modeling capability was shown through weighted scoring models with fixed business thresholds.
Decision logic implementation included Phase-Out Overrides, Low-Base Growth signals, and Strategic Intelligence Flags.
Executive reporting was delivered through a ten-page Executive Summary and a 80+ page Full Report.
Business intelligence skills were demonstrated through KPI design, metric definition, and target setting. Data transformation capability was shown through ERP data cleaning, aggregation, and normalization.
And version control was managed through Git and GitHub.
The platform identified $4.0M+ in annual value opportunities across three areas. Contract negotiation savings were estimated at $2.7M through the Power Tools contract and other high-opportunity products. AP friction reduction was estimated at $0.8M by flagging high-error suppliers like Accenture and CBRE Group for performance improvement plans. Supplier consolidation savings were estimated at $0.5M through consolidation opportunities like Rockwell Automation.
The platform also delivered decision intelligence outcomes: risk mitigation through 18 suppliers flagged for monitoring, contract optimization through prioritized negotiation targets, concentration risk reduction through backup supplier development, process improvement through AP review triggers, and error prevention through the Phase-Out Override.
Three things worked particularly well.
Fixed business thresholds proved more stable than dataset-relative normalization—scores don't change when the dataset changes, making them interpretable and auditable.
Trend analysis emerged as critical to contract decisions—historical spend alone is insufficient, and demand trends reveal products that are effectively dying.
And integration proved powerful—the three engines are more effective together than individually, with consistent scoring philosophy and shared data enabling unified decision-making.
Three things I would do differently next time. I would start with the business question and build the decision engine first, then the dashboard—not the other way around.
I would validate weights with procurement practitioners earlier in the process.
And I would test with real data sooner—synthetic data is useful, but real-world validation would strengthen the model.
Technical improvements include replacing manual weights with machine learning optimization, adding predictive analytics to forecast future supplier risk, automating scoring refresh with real ERP data, and building a natural language interface for conversational querying.
Business improvements include applying the framework to actual procurement data from a real organization, refining weights and thresholds with procurement practitioners, incorporating delivery performance, lead times, and quality metrics, and adding supplier financial health and credit ratings.
Scalability improvements include deploying the platform as a web-based application accessible to procurement teams, designing for multi-tenant architecture with configurable thresholds, connecting directly to ERP systems through API integration, and implementing role-based access for different stakeholder groups.
This project demonstrates how SQL, Tableau, and business logic can be combined to build a decision intelligence platform that helps procurement leaders identify high-risk suppliers requiring intervention, prioritize strategic contracts with clear savings potential, optimize supplier portfolios by mitigating concentration risk, and support executive governance with actionable recommendations.
The platform transforms raw ERP data into actionable intelligence, enabling procurement to move from reactive reporting to proactive decision-making with $4.0M+ in illustrative annual value opportunities.
Prepared By: Oje Ebhota
Date: August 2026
Classification: Public – For Portfolio and Case Study Purposes