Saturday, 26 November 2016 03:52

Searching for relevance


Analyzing keyword searches can determine which searches were most successful in achieving the desired outcome, whether success is defined as finding specific information or improving conversion rates on e-commerce sites. Relevancy tuning helps users zero in on the content they want.

Muck Rack helps communications and public relations professionals find the right journalists and bloggers to pitch to promote their company’s news by searching any keyword, competitor, beat, outlet and media type. In addition, Muck Rack monitors articles about its clients’ companies so they can better respond to news by contacting the journalists creating content.

When the client company is looking for a journalist, effective and tailored search is essential to finding the right match. “We don’t believe in the mass blast,” says Rob Shapiro, director of product strategy at Muck Rack. “Instead of sending a pitch to 5,000 journalists, today’s PR professionals are much better off identifying the five people that have the right content knowledge, experience and audience, then working to build a real relationship with them.”

To achieve that level of granularity, Muck Rack needed not only a robust search engine, but also a software solution that would provide vigilant relevancy tuning to optimize search results for its customers. The company decided to deploy Quepid, an analytics tool from OpenSource Connections. The solution is run on top of Elasticsearch, an open source search tool from Elastic. Muck Rack uses Elasticsearch to search on information collected from journalists’ tweets and social media profiles, articles published online and content shared on social media.

Quepid produces a quantitative relevancy score that can be monitored over time. An initial baseline score is obtained from relevancy ratings by subject matter experts or analytics data. When changes are made in search algorithms that are used to locate journalists, Quepid uses the training set to calculate how far the new results are from the target relevancy level. Subsequent scores are calculated automatically and monitored by the search business to ensure search works right for their users and content.

“Our primary concern is understanding the impact of changes we make on a daily basis,” Shapiro says. “We did not want to inadvertently reduce relevancy, which meant we needed to be able to easily check the results before and after changes. For example, we tweaked our search algorithms to improve negations of certain keywords to prevent results that do not match the user’s intent.” This filters out false hits, such as a journalist who writes about banks (financial institutions) when the target is one who writes about riverbanks (environment).

Maintaining relevancy

The parent company of Muck Rack is Sawhorse Media, which hosts the Shorty Awards for social media influencers. “Our co-founders Gregory Galant and Lee Semel found that journalists were using social media and tools on the Internet more than other early adopters,” Shapiro says. “Although there have always been less specific places to find journalists who covered various topics, there has never been a searchable, constantly updated repository that integrates this many content sources and helps communications professionals identify the most relevant people creating the news.”

Many companies find themselves in a position of needing to change from one search engine to another, and Quepid can be used to ensure that relevancy is being maintained during the transition. “Quepid can track and monitor the new search environment to ensure confidence in the new setup,” says Karen Renshaw, a search relevancy consultant who has experience using Quepid. “Quepid is the only product that allows a query-by-query analysis to compare the old with the new.”

The main goal is to make sure that users see equally relevant results under the new search engine. “Search data can be aggregated to identify the ones that lead to the most profitable outcomes, and then tested on the new system,” Renshaw explains. “Relevancy scores can be recalculated instantly to ensure that the revenue of high-value searches is not disrupted.”

Even when the search environment is not undergoing change, frequent tuning is required, and analytics helps in that process. “Organizations do not always understand that maintaining a search environment is an ongoing process,” Renshaw says. “It is not a plug-and-play situation.” Considerable input involving human action may be required, and this should be guided with support from analytics that verify relevancy.

Quepid was initially developed as an in-house product for OpenSource Connections, which provides professional services in the area of search. After its customers saw the product in action, they wanted to purchase it for their own use. “Increasingly, any search engine starts with an acceptable level of search relevance,” says Doug Turnbull, search relevancy consultant for OpenSource Connections, “but we’re finding organizations increasingly need to customize results, ranking for their content, users and use cases. Google has trained users to rely on search—so controlling relevance ranking, it turns out, is vital to keeping users engaged.”

Achieving better search is not an IT problem, although it is often given to IT to solve. “Developers need the knowledge of SMEs, but traditionally there has not been a good way to share the input,” Turnbull says. “The solution is a test-driven relevancy dashboard used to apply a rigorous process for iterative improvements to search relevance. You need to define what ‘good’ looks like and then measure against that. Quepid provides transparency into search results, based on rigorous testing practices.”

Customer ID and access management

One of the hottest areas of information security is customer identity and access management, as companies strive to protect corporate assets and deliver personalized customer journeys. Gigya offers a SaaS platform that collects first-party data for visitors to websites, unifies it from multiple sources into a customer profile and segments it to maximize the chances of conversion.

“Gigya’s platform is the engine behind some of the world’s largest customer identity and access management systems,”’ says Ian Hametz, documentation team leader for Gigya. “We have developers all over the world who need to access our documentation, but we found that our existing search solution was not providing users with the relevant results they were expecting.” After checking out different options and testing available solutions, Gigya chose Swiftype, a hosted search solution.

The search solution lived up to Gigya’s expectations, particularly the analytics. “Our biggest gain was the analytics,” Hametz says. “We now have much better insight about what our users are searching for, and we are able to use that data to surface relevant content that was otherwise hard to find.” Gigya also finds that Swiftype brings additional value by helping to identify new needs. “We are now creating content to address questions that we might not have thought of on our own,” he adds.

Hametz speaks of the positive relationship his company has with Swiftype, which was an important factor in the decision. “They showed creative thinking when we needed some custom tweaks to get our implementation where we needed it to be,” he explains.

Swiftype was designed with business users in mind. “Often, the business users who are most impacted by the success or failure of search applications provided for their customers are left out of the process when a search application is developed,” says Matt Riley, co-founder and CEO of Swiftype. “We wanted to offer something that had layers for business users so they could see how well search is performing. With Swiftype, stakeholders such as marketers, customer service specialists and others can see how well the search application is performing for their group of customers.”

Companies invest a lot of time and money in creating content and knowledgebases, according to Riley, but customers still have trouble finding what they need. “About 94 percent of searches are three or fewer keywords,” he says. “That does not give a lot of guidance to the search engine to find some very specific information.” In addition, it can be hard to discover what is not happening. “Many companies are not sure what topics they are not covering effectively,” Riley adds. “Our analytics reveal null results and items that users are not clicking on. Those insights can help guide the development of new content or revised navigation.”

Business user access

Sonus Networks, a provider of secure and intelligent cloud communications, was expanding into a broader range of customers and partners as a result of technological advances in its products. The company’s analysts work primarily in the SalesforceService Cloud environment, while customers and partners access information through Sonus’ Global Support Portal. All the groups needed intelligent access to Sonus’ information resources, which were distributed across many systems. To connect across those sources and improve search relevance for its users, Sonus picked Coveo for Salesforce—Service and Community Cloud Editions as its search solution.

Within just a few weeks, Sonus was able to offer its users a cloud solution that provided single sign-on access to the information that was scattered throughout the enterprise. One of the characteristics of Coveo’s solution is accessibility for business users, who can gain insights to searchers’ behavior through analytics and can optimize relevance without the intervention of IT. That ability was vital to the success of its self-service initiatives, because the number of calls and support tickets was growing by 30 percent each month. Sonus reported that it was headed for a 20 percent improvement in self-service resolution rates within the first six months that Coveo was deployed.

“As we moved our software to the cloud, we spoke to support managers using Salesforce Service Cloud to understand their need for analytics,” says Richard Tessier, VP of products at Coveo. “There is a lot to learn about user behavior from the clicks and the path they take on their searches. The valuable resources can be promoted by ratings indicating a particular resource helped the user solve their issue. This feedback can be used to instruct the relevance ratings going forward.”

Coveo can detect when users do not get any results. It can also reveal when a user has to click deeply into a website to find the target information or when many users are seeking information about the same product. “An ongoing process of improving relevancy is needed to keep the search at a high level,” Tessier says. “We use the information gathered in analytics to try to make the process as automated as possible. We have seen major improvements in user metrics once a client has enabled the automated relevance function.”

The future of search will involve a great deal of proactive support from intelligent applications. “A user will be able to load up a case, and the search tool will pick out bits and pieces of the case and then examples of resources that have been used to solve previous similar cases,” Tessier predicts. “The agent will not need to provide keywords—the searches will be triggered around the information available and will provide whatever information is valuable at that time.”

Source : http://www.kmworld.com/

Author : Judith Lamont


World's leading professional association of Internet Research Specialists - We deliver Knowledge, Education, Training, and Certification in the field of Professional Online Research. The AOFIRS is considered a major contributor in improving Web Search Skills and recognizes Online Research work as a full-time occupation for those that use the Internet as their primary source of information.

Get Exclusive Research Tips in Your Inbox

Receive Great tips via email, enter your email to Subscribe.