Streamlining Procurement through AI-Driven Item Data Harmonization

Introduction

In the procurement industry, large organizations often face significant challenges when managing data across multiple Enterprise Resource Planning (ERP) systems. A global company with diverse business units encountered a common yet complex issue: inconsistent item naming across its ERP systems. This lack of standardization hindered the procurement team’s ability to gain a unified view of purchases, leading to inefficiencies, missed opportunities for cost savings, and operational silos. The company sought a solution to harmonize its item data, enabling better spend analysis and procurement consolidation. Oraczen, a leader in AI-driven data solutions, delivered a comprehensive solution to address this challenge, resulting in significant improvements in data consistency and procurement efficiency.

Industry Context

The procurement industry often grapples with data harmonization challenges, particularly for global organizations with decentralized operations. According to a 2024 report by Gartner, over 60% of large enterprises struggle with inconsistent data across ERP systems, leading to an average of 15% overspending due to missed consolidation opportunities. The rise of AI and machine learning has opened new avenues to address these challenges, enabling organizations to achieve greater efficiency and cost savings.

The Problem

Inconsistent Item Data Across ERP Systems

The company operated with multiple ERP systems, each used by different business units with varying naming conventions for the same items. For example, an item might be labeled as "A" in one system, "A*" in another, and "B" in a third, despite being the same product. This inconsistency created several issues:

1. Lack of Visibility:

The procurement team could not accurately track purchases across the organization, making it difficult to identify opportunities for bulk purchasing or supplier consolidation.

2. Inefficient Spend Analysis:

Without a standardized way to categorize items, spend analysis was fragmented, leading to missed cost-saving opportunities.

3. Manual Mapping Challenges:

The company initially considered creating a manual mapping table to align item names (e.g., converting "A*" to "A"). However, this method was labor-intensive, error-prone, and unsustainable as the volume of data grew.

4. Data Discrepancies:

Variations in item descriptions across systems made it nearly impossible to identify adjacencies or related items, further complicating procurement strategies.

This problem, often referred to as Item Data Harmonization, is a common challenge in procurement, especially for organizations with decentralized operations. The company needed a scalable, automated solution to standardize item data, improve visibility, and enable more effective procurement strategies.

The Solution

AI-Driven Data Harmonization and Clustering

Recognizing the limitations of manual mapping, Oraczen delivered an AI-driven approach to harmonize item data. The process was designed to not only standardize item names but also enrich the data, cluster similar items, and provide actionable insights for procurement optimization. The solution was executed in two phases, with the following steps:

Phase 1: Data Enrichment and Clustering

Oraczen began by separating the data into three broad analysis groups based on item categories and characteristics:

1. Group 1: Chemical & Biochemicals

CAS Enrichment with AI: For items in the chemicals category (Level 1, or L1), Oraczen performed Chemical Abstracts Service (CAS) number enrichment. CAS numbers are unique identifiers for chemical substances, enabling precise identification. Enrichment was based on columns such as item description, supplier L1, and manufacturer name. For example, below record is enriched with CAS number.

CAS Not Possible - Notes: For chemicals where CAS numbers could not be assigned (e.g., due to incomplete data), detailed notes were provided to explain the gaps, ensuring transparency.

2. Group 2: OEM Vendors with MPN/VMN

MPN/VMN Enrichment:

For Original Equipment Manufacturer (OEM) items, Oraczen enriched Manufacturer Part Numbers (MPN) and Vendor Material Numbers (VMN) using AI, again based on item description, supplier L1, and manufacturer name. For example, below records are enriched with MPN and VMN details using AI

Validation & Enrichment with Source Citation:

The enriched MPN/VMN data was validated, and source citations were provided to ensure traceability and reliability.

3. Group 3: All Others

Dimensional Enrichment with AI:

For remaining items, Oraczen enriched additional dimensions, including item type, item functionality, broad industrial use, specific molecular or chemical name, and specific technology or activity group. This enrichment was also based on item description, supplier L1, and manufacturer name.

Clustering:

AI-driven clustering was performed to group similar items together, even if their descriptions varied across systems. For example, items labeled "A," "A*," and "B" were clustered if they shared similar characteristics.

Cluster Analysis:

The clusters were analyzed to identify patterns, adjacencies, and opportunities for procurement optimization.

Phase 2: Feedback Loop and Data Standardization

After delivering the enriched and clustered data to the company, the procurement team used Oraczen’s insights to update their source systems. Specifically:

Standardizing Descriptions:

Items identified as similar within clusters were assigned a common description name in the source ERP systems. For example, if "A" and "A*" were clustered together, both were updated to a single, standardized name.

Improving Data Quality:

The company used the clusters to ensure that future data entries followed the standardized naming conventions, reducing discrepancies over time.

Ongoing Maintenance:

The procurement team took ownership of maintaining the standardized data, using the AI-generated clusters as a foundation for continuous improvement.

‍

Technology and Approach

Oraczen leveraged AI and machine learning techniques to automate the harmonization process. Key technologies included:

Natural Language Processing (NLP): To analyze item descriptions and identify similarities, even when naming conventions differed.

Clustering Algorithms: To group similar items based on enriched attributes like CAS numbers, MPN/VMN, and other dimensions.

Data Enrichment Tools: To append missing attributes (e.g., CAS numbers, item functionality) using external databases and AI-driven inference.

The approach was designed to be scalable, allowing the company to apply the same methodology to other item categories or business units in the future.

Results and Benefits

The AI-driven item data harmonization solution delivered by Oraczen provided significant value to the company:

Improved Visibility:

By standardizing item data across ERP systems, the procurement team gained a unified view of purchases, enabling better spend analysis and decision-making.

Reduction in Data Discrepancies:

Post-implementation, the company reduced item naming discrepancies by 85%, as measured by the number of unique item descriptions for identical products across ERP systems.

Cost Savings:

The clustering process identified opportunities for consolidated purchasing. For example, the team discovered that multiple business units were purchasing the same item from different suppliers at varying prices. By consolidating these purchases, the company achieved an estimated 17% efficiency savings in procurement costs.

Time Savings:

The AI-driven solution reduced the time spent on manual data mapping by 90%, freeing up procurement teams to focus on strategic activities like supplier negotiations.

Supplier Consolidation:

The clustering analysis enabled the company to reduce its supplier base for certain item categories by 20%, leading to better pricing leverage and streamlined operations.

Adjacency Insights:

Clustering revealed related items (e.g., items with similar functionality or industrial use), enabling the procurement team to explore new supplier relationships and optimize their supply chain.

Testimonial

A procurement manager at the company shared, “Before this solution, our team spent countless hours trying to manually align item data across systems, often with limited success. The AI-driven approach from Oraczen not only saved us time but also revealed opportunities for cost savings we hadn’t even considered. It’s been a game-changer for our procurement strategy.”

Challenges and Lessons Learned

While the project was successful, a few challenges emerged during implementation:

Data Gaps:

Some items lacked sufficient information for CAS or MPN/VMN enrichment, requiring manual review and documentation (e.g., "CAS not possible - notes").

Stakeholder Buy-In:

Initially, the company was hesitant to move away from a manual mapping approach. Oraczen had to demonstrate the value of AI through early results to gain approval.

Change Management:

Updating source systems required coordination across multiple business units, highlighting the importance of clear communication and training.

Key lessons learned included the need for robust data governance to maintain standardization over time and the value of involving stakeholders early in the process to ensure alignment.

Conclusion

The AI-driven item data harmonization solution developed by Oraczen transformed the company’s procurement process, addressing the critical challenge of inconsistent item naming across ERP systems. By leveraging AI for data enrichment, clustering, and analysis, Oraczen provided a scalable, automated approach to standardize item data, improve visibility, and unlock significant cost savings. The success of this delivered solution demonstrates the power of AI in solving complex procurement challenges, offering a blueprint for other organizations facing similar issues. As the company continues to refine its source systems and explore future enhancements with Oraczen, the foundation laid by this project will enable more efficient, data-driven procurement strategies for years to come.