Keyword search is dead. Embrace machine learning.

Interfaceing

It is about understanding a user's journey and machine learning

Keyword search as a primary signal is dead. Machine learning killed it.

The paradigm of keyword search is over 25 years old. Given the emergence of machine learning and advanced ranking models, it is not the primary source of relevance for insights. Keywords as a driver for information retrieval is outmoded and ineffective. Modern machine learning provides better mechanisms for improving the outcomes of information retrieval requests by providing predictive solutions based on user actions and context. In this article, we will take a high-level look at the how and why we evangelize the death of keyword search. We believe the time for an investment in technology that enables user next action and deeper insights is now.

Why now?

Now is the time to embrace the intersection of technical innovation. Machine learning is ready to deliver on the promise of previous years. Sentiment analysis, text analytics and machine learning provide solutions for use cases ranging from fraud detections to technical support chat bots. Thanks to the productization of the technologies, more complicated use cases are available to most companies.

Accessibility to the technology due to lower cost is important, and users are ready. Technologies like Siri, Alexa and Google Assistance demonstrate some of what is possible. It isn’t about ‘relevant’ results, it is about providing users with the information they need to take the next best action.

A better way

You do not have to look very far to find parallels. There is an established concept in marketing; the customer journey. This approach to planning content delivery to site visitors and leads based on where they are on the customer journey. In marketing, you provide different content or potential interactions depending on where the potential customer is their process to transact.

Understanding how your users use technology and business process is more important than ingesting all your SharePoint content. Mapping how the users will interact with the system and how search can impact these existing business processes should be the first step in developing any new search system, especially if you are replacing a legacy system.

Making the switch

A major takeaway from the customer journey paradigm is focusing on the next best action. What is the next best action for a user, in the specific context? What moves the process forward and reduces user failure at a given task.

Designing search powered applications and services that embrace machine learning and other advanced signals is what we recommend to all of our clients. It is not about simple ‘search’ anymore. You have the ability deliver an organization wide answer engine. The standard use case is to interrupt a potential support user, providing information that allows that user self-service before creating a support ticket. To do this, you need to understand who your users are, what they are looking for and in what context.

Focus on the users and the tasks they are trying to complete in context instead of a presenting a list of ‘relevant’ results. Understand and map the user’s journey and bring insights to various points along their path or business workflow.

Finally, for this to work, it needs not to be opt-in. Ideally, search and insight improvements are integrated deeply. They are less successful when users are required to leave some process to go actively looking for something.

Quick takeaways for embracing insights:

  • The experience needs to consider a user’s point of view, not an internal process or point of view.
  • Design the experience for multiple user touchpoints.
  • Define metrics for success and make sure you are capturing and reporting on them

The time is now

MC+A has over 10 years of experience design and deploying solutions that connect users to existing business intelligence within your organization.  Drop us a note and we can help you get started

Get Started