Let’s Get Personal

I remember becoming interested in joining BA Insight a couple of years ago. At that time, the market was called “Enterprise Search”. In the past few years, both Gartner and Forrester have renamed the category. Gartner calls it Insight Engines and emphasizes the use of relevancy methods and proactive or interactive information delivery in context. Forrester refers to it as Cognitive Search and Knowledge Discovery, highlighting the use of AI technologies such as Natural Language Processing and Machine Learning. Along with this renaming, there have been additional attributes discussed by both analyst firms, which in my opinion, are imported from internet applications including AI, Machine learning and specifically…personalization. Does personalization improve the employee search experience? I think the response is a resounding, “It depends.” Let’s say you have deployed personalization and it is really, really good. Now let’s say one of your team members searches for something that is in another system.  In the results, the employee won’t find the information needed because there is no access to it.  So, do you go and personalize every enterprise and business system that you have? Even if you did, do employees know which system(s) to go to? For personalization to work, the search platform needs to be connected to key business systems. An employee might need to find a contract in Documentum, a product description in SharePoint, and service requests out of ServiceNow in order to do his/her job in a particular week. Without connections to these systems, relevant information will never be found, and the employee will either have to search multiple systems (unlikely) or simply give up (more likely). Do we want personalized and irrelevant? No amount of personalization will help here as the relevant information is not available. Even when there is a connection to other systems, this may not be enough to bring back relevant information. When you have multiple systems, unless you have a strategy to make sure information is tagged properly, then the results will be a keyword search. That is a no-no when it comes to search. The search results will be all over the place and will not meet the user’s expectations. Manual tagging has not worked and it will not work. You must consider ways to put technology in place to meta tag information. As search expert Martin White astutely brings up in a recent article, personalization can be detrimental without proper context and search basics. White describes a common scenario where an employee may move to a different location and department within an organization. How long does it take for search personalization to realize this move (if ever)?  A week? Two weeks? In that time, the employee’s queries will be relevant only to his old position/location, which may cause him to lose trust in the enterprise search solution and stop using it. So why do the majority of enterprise search projects fail? The promises of Cognitive Search or Insight Engines with Artificial Intelligence (AI) and Natural Process Language (NLP) will not be achieved if the basics of connectivity and proper meta-tagging are not in place. The same is true with personalization. There is no short cut to a good search experience.