Cognitive Search and User Intent

What’s in-between your users and a great experience is a search box.

– Michael Cizmar

Cognitive Search – The Forrester Wave update

Last week, Forrester released “The Forrester Wave™: Cognitive Search, Q2 2019.”  Released initially in 2017; this report validates technology vendors across a variety of categories including market presence, strategy, and their current offerings with regards to Cognitive Search.

The report’s positioning is dependent on Forrester’s evaluation of the vendor’s market presence, strategy, and current offering in regards to Cognitive search. To be considered for inclusion in the report, vendors must have a comprehensive, differentiated cognitive search solution, that is not technologically embedded in any particular applications.

The selection process and Forrester’s thoroughness have made the report a benchmark by which many in the technology sector use as a starting point for their technology platform evaluations.  The report’s release also serves as a timestamp by which we can see and track how vendors score against each other and what Forrester views as the market needs and requirements.

We congratulate our partners Attivio, Coveo, Elastic and Lucidworks for their positioning on the Wave report again this year.

Opposite of the vendors is the report itself to consider.  Here we see evolution and reaction to the market. In 2017, Forrester stated, “Relevancy And Completeness Matter Most” (where completeness is the scope of what can be indexed). Their position has evolved this year, stating that AI technologies are the key differentiators and that the critical feature is the system’s ability to “Predict users’ intent to boost relevancy”, which they called “Intent Intelligence”.

We couldn’t agree more.

Understand a user’s intent and direct outcomes

While having a universal index is vital, you no longer can rely on simply that you have an excellent TF-IDF (term frequency–inverse document frequency) based index that returns all of the documents that have term matches.  For example, as an online auto parts wholesaler, it is relatively common to see searches containing VIN or SKU information.  This information is not always present in items that the user is looking to purchase given the legacy approach term matching,

A few weeks ago, at Coveo’s Impact 2019 during the Innovation Showcase, we demonstrated a new solution accelerator MC+A developed that when combined with these technologies is transformational for users and customers.   The process starts by defining a variety of intents and leveraging machine learning to define outcomes.  Example searches can be: “2014 Ford F150 Brake Pad” or “Who do we know at Boeing.”

One of the beauties is that it does not have to be in formal or structured language.  SKUs, RMAs, or model numbers are examples of data that can be easily extracted.  Based on the form of the query, we can understand what the user is looking for and thereby change the behavior of the search.

Query Intent – What are you trying to do

Query Intent Tester:

Directing Outcomes

Understanding the user’s intent allows you to individualize the search experience by utilizing a variety of prefiltering and querying before sending the query to any of the platforms listed in the report (letting our partners’ technology do its magic).  Directing your users to outcomes changes the paradigm of the system from a simple transaction to purposeful experience.  Being more purposeful yields to higher conversations and improved self-service.

Sample In Action

We built a simple demo that takes some parts data and uses our accelerator to use intent detection to prefilter on some parts.  Type in 2014 Audi and see the filters already in action.

Do you want to learn more?  Subscribe to our Newsletter where we share industry news and our insights, or contact us for a demo and quick discussion if this is something that might help out your organization.