Imagine kicking back with some popcorn to watch Netflix and having to scroll through almost 15,000 available titles. Even with a good search engine and well-organized categories, sub-categories, and sub-sub-categories, you’d be likely to face decision paralysis—and finish your popcorn before you’ve picked a flick. Thankfully, Netflix knows what you like. Its smart recommendation system suggests movies and shows based on what you’ve already watched.
Smart recommendation systems have amazing potential for knowledge management. These systems integrate information about users and their past behavior to filter items and predict which will be most relevant or appealing to a particular user. That means they can do two amazing things:
- Improve enterprise search by ensuring the top results are relevant and tailored to individual employees.
- Enable alerts that push content to employees based on who they are, what they’re working on, and their prior interactions with KM systems and other work applications.
At APQC, we’re already seeing organizations use smart recommendation systems to provide faster, easier, and more targeted access to content and experts. Many are using Microsoft Delve (and the underlying Microsoft Office Graph, which aggregates user data across Office 365 apps) to push recommendations to employees based on their current network, what they’re working on, and what their colleagues are doing. For example, an energy utility we spoke with applied smart recommendations to its people profile system, helping employees expand their networks and connect with likeminded colleagues across the company.
But there’s a catch: implementing a smart recommendation system can be a little tricky. From a simplistic perspective, there are two different ways to develop a recommender system:
- Item-based filtering bases predictions on a user’s prior interactions with content and the system.
- User-based filtering bases predictions on user type and characteristics.
Which approach will prove easier to implement and more effective in the long run depends on several factors including the number of end users, the frequency with which new content is added, and the diversity and number of people in different roles. More sophisticated recommendation engines combine both approaches to generate highly tailored recommendations.
Additionally, smart recommendation systems often require integration across multiple platforms—including those outside KM’s control, such as HR and customer data systems. As a result, technology integration, security, and privacy concerns may arise. And some early adopters have found that the algorithms in off-the-shelf platforms cannot be sufficiently customized to provide accurate recommendations.
Still, implementing a smart recommendation system for KM may be well worth the hassle, especially for large organizations with many employees, clearly defined roles, and lots of knowledge content. These systems might prove to be key to achieving that ultimate KM goal: delivering the right knowledge, at the right time, to the right people.
To learn more about how smart recommendation systems, chatbots, cognitive search, virtual reality and other leading-edge tech can benefit KM, see APQC’s new collection, Emerging Technologies for Knowledge Management.