Common Searches

Relativity Search

Enabling users to be amazing at their jobs

Relativity Search

Search is one of the most critical yet least intuitive aspects of Relativity. It is a gatekeeper—an obstacle to end users doing their job effectively. In 2020 we set out to transform Search into a product that enhances users' expertise and extends their abilities, enabling them to be amazing at their jobs.

Common Searches


After working for several years on Analytics, I was provided an opportunity to switch products and expand my role. In July of 2020 I joined the Search vertical where I have helped build the team, define the vision, and influence the roadmap. The foundations of this vision are:

  1. Powerhouse List Page - Update the table component from a static display of information to a dynamic workflow-driver

  2. Simply Powerful Search - Reduce the barrier to entry and increase confidence by streamlining core workflows, seamlessly integrating advanced functionality, and humanizing querying language

  3. Instant Insights - Extend users' abilities by leveraging AI and ML to surface key insights and provide contextualized assistance

The project is still in the early stages, but it is some of the work I'm most proud of and the early results are very promising. This work is the foundation for massive improvements to Search in 2021 and beyond.

Our users shouldn't need an advanced degree to search. Instead, Search should extend their expertise and enable them to be incredible at their jobs.

Aspirational to Tactical

As we defined the multi-year vision for Search, we identified the three high-impact problem areas to focus on in 2021. Addressing these areas provide a step function improvement for the user experience and set us up for the future-looking work we have planned. They connect the dots from the software today to the vision for tomorrow.

The Vision for Tomorrow

The Problems to Solve in 2020

  1. The list page is a mess and is hard to navigate
  2. The table widget lacks fundamental features
  3. Searching is confusing and click-intensive

This case study is focused specifically on problem 3. The Powerhouse List Page case study focuses on problems 1 and 2.

Hypotheses-driven Design

From the outset I approached this work like a science project. Using user data and feedback as a foundation, I framed the problems and formed hypotheses to improve the experience. Based on these hypotheses I am creating design explorations and working with the UX Research team to to test them, iterating until we've validated an optimal solution.

Centering the User

My top priority throughout this project has been to ensure that the user is at the center of everything we do. Every problem we solve is framed by real world use cases, explorations are informed and scrutinized by internal and external experts, and solutions are tested and validated by the people who use them.

To get the volume of user input we need for a project of this size—one spanning 12+ months—we decided to form a user advisory board. This board is comprised of a variety of user types across key customer segments, all of whom have committed to:

  • A monthly, hour-long call to review priorities, gather feedback, demo work, etc.
  • Ad hoc deep-dives (contextual inquiries, concept testing, usability testing, etc.)
  • Beta testing new features and providing initial feedback before launch

Advisory Board Demographic Breakdown

46+ users

representing 3 primary user types

15 orgs

of varying size & sophistication

5 segments

identified as high priority

User Types

The landscape of Relativity users is highly varied and diverse, but for our purposes the user types can be broken down into three major buckets across two axes: level of sophistication and type of work.

Common Searches
  1. Administrators - highly sophisticated; creating the things to drive workflows
  2. Case Strategists - highly sophisticated; defining workflows, case strategy, QC
  3. Reviewers - less sophisticated; executing workflows, reviewing documents

These three user types are highly representative of the majority of Relativity users and serve as useful context when exploring new features.

Search is confusing and click-intensive

Through a combination of user interviews, quantitative usage metrics, and anecdotal feedback we identified the biggest painpoints with Search:

  1. Too Clicky - Common searches are laborious to create
  2. Too Hidden - Globally useful features are considered "advanced" because they are hard to find
  3. Poor Grouping - Search features are split across 9 different places on the UI

Too Clicky

Throughout all of our user interviews, the single piece of feedback we heard most was "creating searches is too clicky!" As we've begun to dive into this problem—interviewing users, reviewing search composition—a few patterns have quickly emerged:

  • The vast majority of search conditions are looking for people or dates
  • The majority of the rest tend to be a small group of review fields specific to that case

We have come to refer to this as the "people and dates problem." To do something as simple as searching for a specific person takes a minimum of 9 clicks. Add a congested interface into the mix and the task is onerous.

Search Features


  • The majority of searches conditions fall into one of three buckets
  • Field specificity adds work and complexity but not value
  • The bulk of the work is navigating past unwanted conditions

Too Hidden

Another common piece of feedback we received is that features like Keyword Expansion—use AI to surface conceptually similar search terms—have broad appeal but are almost never used because they can't be found. Some have gone as far as to say that Keyword Expansion should be used in every review, but a quick look at the usage data makes it clear that isn't the case.

Search Usage


  • "Advanced" search feature usage is a fraction of a percent of text searching
  • The majority of usage comes from a small subset of users

Poor Grouping

An audit of search functionality on the List Page provided vivid clarity around our users' grouping complaints—by default there are nine different areas containing search-related features. To make matters worse, they are scattered throughout the interface without much proximity or grouping. In this case, a picture says a thousand words.

Clickstream Analysis


  • The are as many as six different features that effectively filter the result set, but no way to see them all in a single location
  • Connected actions—e.g., adding and removing search criteria—aren't always connected on the UI
  • Duplicate search conditions can be added from different locations


Our early research has generated a few key hypotheses that are already yielding promising results in early explorations:

  • Grouping search conditions into common types reduces hunting and increases efficiency
  • Suggesting commonly or recently used conditions reduces clicks
  • Consolidating search features increases comprehension and reduces mouse travel


Using the above hypotheses as a guide, this early exploration reimagines what Search could be like if it were consolidated and its actions streamlined. The result is a dramatic reduction in clicks to achieve common tasks, a centralized location for related functionality, and recouped space across the interface.

Reduced Clicks on Key Search Workflows

-33% clicks

to search for dates and fields

-56% clicks

to search for people and entities

-83% clicks

to use advanced search operations

Next Steps

Coming up next for this project is continued exploration and the typical prototype-test-refine loop as we strive to simplify and optimize search. We have already laid the groundwork for these features with the release of Recent Searches. Looking beyond, we have some exciting ideas to explore utilizing machine learning and natural language processing to make searching more accessible, powerful, and humane.

Freeze Columns Animation
Freeze Columns Explorations
Freeze Columns Explorations

Powerhouse List Page

Case Study (2021)

The list page is the backbone of the Relativity platform. It is the base from which everything else is built, yet outdated functionality and poor organization keep it stuck as a merely an entry point. In 2020, we set out to transform the list from a static display of information to a high performing workflow-driver.

Check it Out
Common Searches