Anand Chakravarty & John Guthrie, Microsoft Corporation
With a still growing number of users and an ever-growing volume of information available, the Internet presents many interesting and continuously morphing challenges in the area of Search. In addition to the scale of data to be processed, the results of user search queries are expected to be highly relevant and delivered speedily. Presenting these results to the user in a manner that enables them to decide and act based on the information provided is a primary characteristic of a good Search engine. Measuring the accuracy and relevance of Search results is thus an important area in the Testing of Search engines. Considering the high volume of information to be processed, the extremely diverse nature of query-intent and the growing expectation of high relevance from Search users, testing of Search engines requires the traditional QA characteristic of passion for quality, combined with a high-level of comfort with ambiguity and strong automation skills.
When we consider the area of search queries that have a location specific intent, there are other factors that become important along with the usual technical problems. Most queries that have local intent have a higher degree of immediacy in terms of the user’s intent to act on the results returned to their queries. There is thus greater expectation of the results being fresh and accurate. With the growing market for Mobile devices, searches with local-intent are becoming more popular. As a tester, when presented with measuring how well a Search engine performs for such queries, it is important to understand the scale and variety of queries involved. Because there is a high level of ambiguity and variation in search queries and results, statistical metrics are a natural tool to measure Search engine quality.
In this paper we cover the methods used to obtain metrics for measuring relevance of search queries with local intent. Testing is done while fundamental components are in a dynamic state: rankers, intent detectors, content, location identifiers, etc. A QA team that comes up with good metrics to measure the quality of search results in such scenarios increases the quality of Search Experience delivered to their users, and helps to evolve the quality of solutions implemented in this extremely challenging problem space.