International integrated energy company with interests in exploration, production, refining and supply

Purpose of the project

-	Extraction of Attributes to enable Parametric Search
-	Cleansing and Normalizing of Inventory data into a transactable format

Volume

-	700,000 Items

Type of items

-	MRO Items related to parts of Plant and Machinery, e.g. Turbines, Pumps, equipment, parts, PVF

Process

-	All items were classified to a customer taxonomy. Domain experts suggested new categories and attributes for items which did not have a relevant class in the existing schema.
-	Data was cleansed and normalized, incl. Mfg. Name and Supplier Name.
-	Mandatory attribute values such as Manufacturer Name, Mfg. Part No., Supplier Name, Supplier Part No., were extracted from Description.
-	Similarly, product specific attributes were extracted to templates provided by the customer. Item Descriptions were enriched by visiting Manufacturer Websites and sourcing additional information for Customer distributed items.
-	Similarly, product specific attributes were extracted to templates provided by the customer.
-	Identified the duplicate items sent by the same supplier, and then allowed the supplier to select the best described item and hence created a Master Catalog. Duplicates were identified by Mfg. Name, Part No. and Item Description.
-	Quality Control and Quality Assurance.

Key Success Factors

-	Domain Expertise coming from experienced professionals from Mechanical and Electrical Industry.
-	Dedicated team of 120 professionals generating 80,000 items per month.
-	Strong process definition and on-site project manager.
-	Ability to handle large datasets, and management of production facilities using Unilog's proprietary software Applications.