Entity search techniques for expediting entitlement resolution in technology support services

Abstract

'Entity search' addresses the problem of finding data objects most accurately identified by multidimensional attributes, where individual attributes span multiple data sources with duplicate or inconsistent values. We have applied efficient entity search techniques to expedite entitlement validation in IBM's Technology Support Services (TSS) division. Entitlement validation enables a customer entity to obtain services - e.g., repair or replacement of a product entity - under warranty or contract. Expedited verification ensures customer satisfaction but is challenging when a single critical customer-provided attribute error results in a validation failure, and resolution requires exploring a large search space to find alternate entitlements that are most similar to the original request to be viable alternatives. We have built an entity search tool (consisting of data field interpretation, approximate matching, and weighted scoring) that has accelerated alternate entitlement search in TSS from minutes to seconds. Core concepts include the use of error correction/exploration techniques (where different entity identity attributes are varied based on nearness measures), entitlement repositories searched using the variations, and attribute-level fuzzy matching used to compute similarity scores between the original request and alternate entitlements and return the highest scoring results. Our experiences with deploying these tools in a large worldwide organization are described.