I’ve been working on an interesting project with a higher education institution. They have Oracle Service Cloud Enterprise, which ships Knowledge Foundation as standard. Knowledge Foundation is what used to be known as RightNow Answers. They came to me with a need to fix their knowledge base—the search returned irrelevant results, there was a lot of outdated and incorrect answers, and there hadn’t been a knowledge manager role in the organisation for more than a year. The platform was good enough to do the job, but their previous workflow had a built-in bottleneck and the state of the knowledge base had only gotten worse since that one person left.
Knowledge Foundation allows you to serve knowledge articles to different audiences via interfaces and respective access levels. My client had three interfaces, each with a matching access level, set up to serve separate and distinct audiences—current students, prospective students, and internal staff. But, with an unclear knowledge strategy, content relevant to one audience was often appearing in more than one access level, making the whole experience of using search a difficult one, no matter the audience. There were also a few basic features that could be enabled, and some minor configuration changes that could be made, all of which contribute to a far more effective knowledge base. I’m going to share the steps I took so you can improve the search effectiveness of your Oracle Service Cloud knowledge base. For now, I’ll leave the roles and workflow process for another post.
1. Know the reason your organisation may have more than one interface and/or access level, and clearly communicate the differences between them. In my client’s case, when a staff member searched for information to do with lecture recordings, for example, they were getting a lot of answers relevant to students. That results in frustration and disillusionment with the knowledge base. I bulk-edited answers to update them with the appropriate access level.
2. Enable Search Result Limiting. How many pages get returned when you use a common search phrase? If you’re getting pages and pages of results, many of which are irrelevant, it’s worthwhile going into the configuration settings to tune the search results. Search Result Limiting uses an AND search except where there are no answers, then it falls back to an OR search. As an example, students like to know when the coming “census date” is. This configuration setting took 60+ results of “census” or “date” down to six relevant results of “census” and “date”.
3. Oracle doesn’t use keywords in the same way that you probably do. When you put key phrases into the keywords field of an answer, it artificially boosts that answer’s weighting in search results when that term is used. Most of us use keywords as synonyms, but Oracle provide a text file to do this job. Monitor the Keyword Searches report and add commonly occurring synonyms to aliases.txt in the File Manager of your configuration settings. Keep the keywords field blank unless necessary.
4. Enable SmartAssistant Auto Tuner across all your interfaces. I don’t know why this feature isn’t turned on by default, because it’s so helpful to search effectiveness. The Auto Tuner is continually learning answer relevancy and makes adjustments automatically. It’s influenced by how often agents reuse answers in response to incidents and will push those most reused answers, thereby deemed most relevant, higher up the results in both customer portal searches and the Smart Assistant suggested answers in the agent console. When you click on SA Auto Tuner in the config settings, you see a bunch of weightings under the current search configuration, and what the suggested config would be for a tuned search config. From here you only need to click one button “Accept new config” to have those suggestions applied. Once a week, a new datapoint is collected for the search relevancy graph that is also on this page. It’s early days for my client, but I suspect that the more the agents interact with SmartAssistant and Best Answer features in their ribbon, the higher these relevancy percentages will go. I’ll have to revisit this theory later.
5. Ensure you enable the Best Answer button in your agents’ workspace. Related to the previous step, the Best Answer button allows agents to select the best answer from those that were reused in the incident reply. This is a significant input to the SmartAssistant Auto Tuner algorithm and helps that work more effectively.
Separately from Knowledge Foundation, but foundational to knowledge management in general, is to embed a Search First culture. If your OSC agents aren’t using the search features within the console, you won’t see the productivity benefits available with the platform, so don’t skimp on the communications and training.