Allstate Virtual Assist
IOS, ANDROID + WEB
A video chat app for insurance companies to assess auto & property damage
MY ROLE
Product designer
Duration
May 2016 - May 2017
Users
12k+ users, 2k+ enterprise users
Outcomes
90% reduction in process time
A beloved project
In 2016, I worked as a product designer for Allstate and became the lead product designer for Virtual Assist. It was my first role with such a large responsibility for the end-to-end design of a product.
With nothing but an accounting degree and an extremely quiet voice during presentations, I started off at Allstate with minimal design skills, a lack of product sense, and what I called “coding skills” (I meant HTML/CSS). In a year, this project and its team taught me the essentials of functional UX/UI design; lean, test-driven development; open and honest cross-functional collaboration; and the value of shipping usable, beautiful product that benefits users’ lives as much as companies. It has a permanent place of honor on my portfolio.
About the app
VA is a video chat app that connects insurance adjusters with customers and repair partners to inspect damage to cars, houses, and other property. Such repair partners include car repair shops, mechanics, and home repair companies and contractors.
Major outcomes
Virtual Assist took 6 months of design and development to deliver an MVP to production. The MVP consisted of a native iOS app (for repair facilities) and a desktop web app (for Allstate employees to answer calls).
When it was first released, only a few repair facilities in the state of Illinois were calling in to two Allstate employees. After a year in production and constant iteration, Virtual Assist had more than 11,000 repair facilities calling in to hundreds of Allstate employees.
The product was later expanded to include a native Android app and mobile React/Java web app, and used for end customer calls, as well. Using Virtual Assist, Allstate’s estimating process went from an average of 1 week to 45 minutes. Based on its success, Allstate restructured its location and hiring process around Virtual Assist, opening 6 operation centers nationwide to answer and monitor calls.
About the app
Like many insurance companies, Allstate has a complicated claims process. When customers file an auto or property claim, insurance companies will inspect the car or house and estimate the cost of repair. Estimating is a large part of many of these companies’ claims processes, and it largely occurs by sending a company employee (an “adjuster”) to physically visit the damaged asset.
The adjuster may check the same house or car several times as they are being repaired, to make sure the correct repair costs have been reported.
This process is lengthy and costly, and involves sending an adjuster to visit the damaged car or house between 1-5 times. This can lengthen the time it takes to complete a claim by more than a week each time.
Main problems
The estimating process is inefficient and costly, due a lengthy and manual process. This impacts the entire claims process.
- manual process involving travel → high cost, lengthens overall claims process
- clients must wait longer for repairs to be finished → bad user experience
Discovery & research
An Agile XP product team was created, with a rotating team of one product manager, eight developers, and one product designer. The team was tasked to reduce the inefficiencies in the estimating process by the Allstate business.
When I rolled onto the team in the product designer role, I conducted foundational research to understand how the repair process within claims worked. My initial research focused on auto claims for our MVP. I identified major pains and inefficiencies in the process by interviewing both sides of the process - the insurance adjusters, and the shops and partners who conducted repairs.
user interviews with 10+ insurance adjusters
contextual inquiry with 25+ car repair shops
Findings
Two major problems from both sides (Allstate and repair facilities) stood out to the team. The same problem appeared on both sides: both Allstate and repair facilities felt that the process was painfully long and costly. Both parties felt that this negatively impacted their business and relationships with their customers.
adjusters need to contact the repair shops (and vice versa) multiple times throughout the repair process
even a small question often requires an adjuster to travel to the repair shop to inspect the car and ask questions
the repair shop has to wait for the adjuster to visit and approve further repairs; this lengthens the process
Ideation phase
While the problem seemed simple (high inefficiency), the solution was far from simple. It was limited by many key facts discovered during the research phase.
adjusters need specific information that only applies to insurance; they need to talk to the repair shop
adjusters are rarely able to produce an accurate estimate by just looking at photos of the damaged car
a repair partner needs an adjuster’s guidance to provide the specific information the adjuster is looking for
The team ideated, tested, and iterated on a variety of different ideas that could solve these problems while accounting for these limitations.
We decided that MVP would focus on the relationship between repair facilities, mitigation vendors (the equivalent of repair facilities in property), and Allstate. The first feature set would be based on the fact that repair facilities would place a video call, which would be answered by an on-demand adjuster.
Design & user testing
The app’s design and feature set iterated together several times, based on changing requirements and user testing. The design was worked on by various rotating engineers and designers, including myself.
The first iteration was extremely simple, using Allstate branding and a linear flow to let users place a video call. It was a team decision from all three disciplines to keep the feature set and flow extremely minimal until we understood how users would respond to it.
Iterations
I later designed the Android mobile app and mobile web app (for consumers) with the same feature set, which was released while the team refactored the desktop and iOS application. At this point, I became the lead product designer for the app, and was not rotated out for other designers.
The later iterations included a more advanced onboarding process, which walked repair facilities and mitigation vendors through inputting their information. This information would be saved when the user returned to the app, helping connect their call to the correct adjuster. In previous iterations, user information was not saved, which I surfaced as a user pain to prioritize in our backlog.
Allstate colors and logos were removed in order to white-label the product and sell it as a service to others. Other companies, from repair facilities to other insurance companies, had expressed interest in using the app for themselves.
Continuous user research & design
I observed the apps being used in production by our users, and also ran continuous user testing for new features. Running user tests and observing real user behavior always generated unexpected results. I observed users opting to use tablets instead of mobile phones, or using the video feature to discuss paper invoices instead of sending an email. Each round of research resulted in iterations in the product’s design.
Development & implementation
I regularly paired with the XP full-stack developers to understand technical limitations and unblock front-end stories, while coaching them in SASS and CSS. Understanding the basics of our codebase helped me prioritize the backlog when I paired with the PM, as well as identify design bugs in the production software. By pairing with the devs regularly, I was able to initiate other XP product teams to move towards using CSS compilers, recommending SASS as the industry standard.
I became the representative for product design in the Allstate claims space, regularly demoing the product to stakeholders and potential clients with the PM and dev anchor. I became an architect for Virtual Assist’s future, which transitioned into architecting the claims space’s overall strategy. I was also a facilitator of the difficult discussions of its future (re-inceptions, scopings, iteration decisions, and kickoffs), where I recommended evolving our approach to user research to become more quantitative.
Retrospective
Virtual Assist was one of Allstate’s most-used apps, and its design was still evolving and changing when I left. Looking back on its more humble beginnings, its unique problem set and rapid expansion offered a lot of challenges for me as a designer.
One of the most challenging aspects of the product’s design was balancing design debt with immediate user value - a totally new problem for a learning designer trained on agile and lean, like myself. As a team, we had chosen to focus on designing for the user and business problems which we could solve in the next few months - not the next year.
This mindset, while it drove the product’s success, also lent itself heavily to design debt. User testing was able to inform when and why we should focus on long-term or short-term solutioning; as the product was iteratively designed and then released, I conducted regular user observations and user tests over the course of six months. Whenever I brought back poor usability as a finding of testing/observation for a short-term solution, the team knew a longer-term solution was the better path. This valuable learning has stayed with me for the rest of my career.