PERFECTING THE SEARCH
Keep it Simple or Complex? An Enterprise Software Dilemma
As LexisNexis's product line for enterprise legal management software, CounselLink provides a platform for thousands of professionals to review daily e-billings from their legal vendors. Many users rely on the software’s search functionality to find and review important billing information amongst hundreds of line items. However, as this is a legacy product that was built 16 years ago, the search experience now felt slow and difficult to use.
I was the design lead tasked with improving the search experience for charge items under individual invoices and re-modernizing this critical component of the app. The project was awarded "Show Me the Data" Pendo award in 2018 for using analytics for making better design decisions.
I have omitted confidential information in this case study to comply with my non-disclosure agreement. All information reflects my own opinion and does not necessarily indicate the views of the employer.
Not only was the search performance slow, the old interface also forced users to choose a particular table field before they could start searching the content. This caused many obstacles for the average user, including:
Users didn’t know what table field the search keywords belong to.
The interface required extra clicks on dropdown menus and cognitive load before users could conduct a search.
There was no flexibility in search results if user happens to chose the wrong field.
I led the design of the search experience and collaborated with a front-end developer to quickly prototype and test the design iterations. In addition, I worked alongside a Product Manager to define feature requirements and success criteria.
The old charges search interface
The newly designed charges search interface by default
While the development team was solving the performance issue by migrating invoices data from Sybase to SQL, I started chatting with some of the invoice reviewers who heavily rely on search to scrutinize billing in CounselLink. Not surprisingly, lots of them expected the software to “search like google”. The users wanted the ability to simply type in a keyword, hit the search button and see the results magically unfold. With that simplicity in mind, we decided to replace the old search functionality with a generic search box for any keywords. Fast and simple were the two main goals of this redesign.
With these goals in mind, I started testing this simplified search concept with an invoice reviewers user group. While most of these users loved the simplicity of my approach, a few started asking questions about more complicated use scenarios.
“What if I want to search for charges exceeding $100?”
“How could I make it only return charges containing 10 units, not charges of $10”
“Can I combine a bunch of criteria? Such as charges on this specific date with this timekeeper?”
It was apparent that while the simple search approach gives users a quick and easy start, it falls short for more complex enterprise workflow needs.
To evaluate the extent of this issues, I needed to determine how many users had a need for this more advanced search capability To answer this question, I leveraged Pendo for more accurate analytics.
Heat map of Charges tab and search features from Pendo in a 30 day time frame
Because users were forced to choose a table field to search with in the old interface, there was no easy way to identify their intentions; be it simple or complex search. My hypothesis was that if a user chose the default search qualifier “contains” instead of another one, such as “greater than” or “not equal to”, I would consider that their needs could be satisfied by a simple keywords search. So, I tagged individual search elements on the page to track this usage. After 30 days the data showed that approximately 73% users did not defer to the more complexed search qualifier. So, we could estimate that 7 out of 10 users only needed simple search functionality. Hoever, a solution that only meets the needs of 70% of users was out of the question? So, the questions became, how can we design a solution that neatly balances between light weight searches and searches that are more complex?
Reframe the Problem
Knowing that my initial goal of “fast and simple” was no longer valid, I decided to re-defined the project goal at a high level:
Make the product fast and easy to use for everyone. (again, not just the 73% who use simple search)
Give invoice reviewers more control over their searches.
The first idea that came to my mind was to add an “advanced search” option next to the simple search box. This would keep the initial simple approach, but also give the other small user group a way to add more complicated search criteria. Also, given that the “advanced search” model permeates many consumer sites, it would feel natural and intuitive for the average users to discover and adapt to this feature. Of course, the downside of this approach is that advanced users would need to be redirected to a different screen, which adds extra steps.
What’s more, even though the old search function was clunky and unnecessarily complex, a large number of our users had been employing it on a daily basis for over 10 years. They were familiar with how it worked and reluctant to face drastic changes. I began to consider if it was possible to reconcile the old search pattern with the new simple approach. While keeping the three boxes structure (field, qualifier and search box), I made the default settings for the first two elements to “All fields” “contains”. Because these two elements are already pre-filled for a general search, users could simply type in any keywords to start a quick search, just like how they search on Google. When the results are returned, a criteria number count would displayed next to search. If a user is interested in performing more advanced search, he or she could click on the link, which will open a popup modal and carry over the first search criteria they had just conducted. In this way, users with more complex needs can continue to refine and add onto their search. There are three main benefits in this hybrid approach:
It reserved the old user behavior, thus reducing the learning curve for the new interface.
It enabled a quick-start functionality by pre-filling the two fields for a general search for users with simple search goals.
The criteria editing screen allowed for combined search criteria, which wasn’t available from the old interface.
I understood that ideation was only one step in the full UX design process; once ideas are generated, separate analysis had to follow to decide which ideas to pursue.
I used inVision to build two quick prototypes for each approach and tested them with our customers. As invoice review can be more complicated and demanding in large corporations and insurance sections, so I particularly chose three invoice reviewers from small companies and three from the insurance industry. I also interviewed them individually afterwards for context. While neither of the design approaches showed usability issues, 4 out of 6 customers explicitly showed their preference towards the hybrid model (3 out of 3 from the insurance sector). Their comments also proved my original hypothesis,; that some of our user base would see anything but the most minimal change as a disruption to their daily routine work. These users are only willing to accept small changes that can make their work easier, but nothing too difficult to learn.
The newly designed charges search interface by default (Top), by simple keywords (Middle) and by using combined search criteria (Bottom)
The team coded and released the new search feature once the development team had finished database migration. The lean UX approach allowed us to quickly complete this design within two sprints. Since then, I’ve heard great feedback from numerous users. Looking at the data two months of the release, interestingly, the percentage of users who leverage more complex search criteria actually climbed to 46%. Some of the users even started using this feature for other purposes, such as generating their own charge reports. One user told me that she loved the new search functionality and would often use it to quickly run a few criteria if her department needed to know how many redundant charges are from a particular timekeeper on the same task.
When I look back, there are many valuable lessons can be learned from this project. However, the number one takeaway has to be to always supplement user verbal feedback with actual behavior.
Anthropologist Margaret Mead once said, "What people say, what people do, and what they say they do are entirely different things.”
This proved to be true in this project with my discoveries surrounding use scenarios in invoice charges searching. Our team almost made the incorrect decision to abandon the complex search functionality simply because we kept hearing some users tell us that they wanted something just like Google. Fortunately, citing actual usage data, we pivoted from the original design and created a solution that met the needs of 100% of users.
A second valuable takeaway is that a function can be only as good as how the user uses it and a nicely designed feature can go even beyond its original intent.
In this case, once the design for advanced searching was approved, not only did our users continue to rely on it to find critical information more often, they also discovered a potential use for it to gather charges data for simple reports.