Background
The search bar is the the most common single action after logging in, and the feature boasts 6.23m clicks in 30 days.
Through user recordings obtained through our product analytics tool, Pendo, I identified an common flow that was unoptimized, and slow. Often a customer will perform a search, realize what they’re searching for doesn’t exist, and navigate away to a different part of the application to create it.
This single flow, repeated hundreds of thousands of times a day, was the inspiration behind QuickLaunch - an action centric, machine learning driven launcher.
Research and Process
As the leader of the QuickLaunch project from start to finish, I led the design and a team of developers towards an innovative solution to enhance our searching experience.
I performed user interviews where the user would guide me through their workflow, and some common pain points were quickly identified.
Simpro Premium [Cloud Based Desktop Software]
Team Lead [5 Developers]
User Research
UI Design
1 Sprint [Two Weeks]
Field Service Management
Pain Points
“I’ve just created or edited something, and I’ve forgotten where it is.”
“I have to scroll for miles to find a job that I access all the time.”
“If I can’t find [what I need] I have to navigate somewhere else in the app and create it.”
Our identified pain points were validated and supported through secondary research.
Technical Inspiration
A lot of my initial inspiration for the predictive behaviors present in the final product was driven by the concept of word trees.
A word tree depicts multiple parallel sequences of words. It could be used to show which words most often follow or precede a target word (e.g., "Cats are...").
This concept was extended upon to show likely suggestions for the next word in a users command. The data was sourced from all our users across Simpro - if after a user types “Create”, “Job” is the highest scoring hit then “Job” will be suggested in the search bar, and the user can simply press tab to autofill the command “Create Job”.
Word Tree Example
Designs
After our research concluded, the scope of QuickLaunch included several key features which I developed into high-fidelity designs to test with our userbase.
Simplified mini-workflows
Step-by-step processes for creating jobs, quotes, and invoices, with prefilled data.
Natural language processing
Ability to interpret commands like "Run X report on Y customer from dates X to Y on this cost center."
Show unpaid invoices for DSM Offices
|
Invoices
DSM Offices
Status: Unpaid
Press Enter to finish
Predictive action suggestions
QuickLaunch identifies patterns, such as addresses, and suggests actions like creating a site.
123 Main St
|
Recent item access
Quick access to recently opened items within the application. We enabled this as part of our default open state, and this was a big timesaver for our users.
Previously there was nowhere in our application to access recently opened items.
Search or select action...
|
Billy Brown
Recent Customer 2
Recent Invoice
Machine learning
Enhancing result accuracy based on user activity appwide.