Opendoor trade partner work orders
Opendoor is an iBuyer that buys and sells homes by using technology to streamline the complicated real estate process.
​
After Opendoor acquires homes, the homes usually need some repairs, renovations, or maintenance done so they can be resold. Opendoor Home Operation Coordinators assign work orders to local trade partners (e.g. plumbers, painters, general contractors) to handle these jobs.
Timeframe:
3 months
My Role:
Designer, researcher
My process
Define
User Interviews
Ideation workshop
Design
Early variations
High fidelity mocks
Validate
Design demos
Pilot testing
Refine
Feedback gathering
Design iterations
Problem definition
I interviewed several Home Operation Coordinators, and discovered they rely on tribal knowledge and memory to assign trade partners to do work orders. They also rely on a small percentage of trade partners to do a majority of the work.
The inefficiency was directly tied to financial losses as homes sat in inventory longer and contributed to the following issues:
Uneven trade partner utilization
Some had too much work, some had no work at all.
Inability to scale operations
Over-utilized partners can't keep up with increased work.
Longer time to list homes
The longer it takes, the more Opendoor pays holding costs.
How Might We...
maximize trade partner utilization to list houses on the market faster so Opendoor can reduce holding costs and scale its operations to handle larger inventory?
Ideation workshop
I presented the “current state” end-to-end journey to stakeholders and product partners and facilitated a workshop to come up ideas on how to address the challenges.
Updating trade partner info
There were two phases of the design work. The first phase was addressing a “trash in, trash out” issue where inaccurate data was being stored, so work orders weren’t being assigned appropriately.
I redesigned the trade partner dashboard so operators could update our trade partners’ service areas and other information.
Trade partner matching
Next, I designed an automated system to match trade partners to work orders systematically. We worked on an algorithm to create a “match score” based on the below three criteria:
A great match: a trade partner who fits all the requirements and is not being utilized.
.png)
A poor match: a trade partner who might fit requirements but is over their work capacity.
.png)
User testing and pilot test
I tested the design proposals with Home Operations Coordinators across different geographic markets to validate the design direction. After user validation, we ran a pilot test with real data and actual work orders in Tampa, FL.
​
Through the pilot test, we discovered three issues with the design:
The system is suggesting trade partners based on data we’ve provided, but the data isn’t clean. We must update our information in the system.”
"
— Home Operations Coordinator
Now work is TOO evenly distributed. We want to strategically distribute work to build certain relationships with trade partners.”
"
— Business Operations Manager​
The trade partner I want to pick is not always being recommended by the system."
"
— Home Operations Coordinator
Iterations based on feedback
In response to feedback that the system worked too well distributing data, we added a trade partner classification that would contribute to the match score.
Partner
A trusted trade partner we should assign regular work to in order to retain them.
Rookie
A new trade partner we assign certain work orders to so we can assess their work quality.
In response to feedback that some trade partner favorites weren’t being recommended by the system, I added an option to see trade partners with lower match scores so they could find their favorite.
However, I made it clear in the UI that they would be less preferred matches to discourage favoritism.
Project outcomes
The launch of this new trade partner matching system brought steady increase in trade partner utilization across every market, and brought about the following outcomes:
Stronger partner relations
Trade partners are happier when they can depend on regular work
Increased work capacity
Distributing work increased overall work capacity by 20%
Increased automation
Using algorithms set the team's sights on more automation
