Salman Jamali, Head of Engineering, Strategic Initiatives at Opendoor
A home inspection is one of the final steps in selling your home, and it can take anywhere from 2-4 hours. The process can be inconvenient, uncertain, and stressful. Our goal was to make the process significantly faster and more convenient, providing a fully self-serve, digital and asynchronous experience for our customers.
In 2020, we adjusted our interior home assessment and introduced a virtual experience-using video to make it easier for customers, while preserving the quality of our assessments. In 2021, we decided to take it one step further. Last month, we introduced self-guided home walkthroughs and have significantly reduced the hassle and time associated with in-person interior inspections, while maintaining the thoroughness and detail of our assessments as before.
So, what's happening behind the scenes to bring this experience to life? Platformization. Backed by a strong product-customer fit and a continued focus on unlocking future potential, it is increasingly important for us and similar companies in the digital space.
At Opendoor, we used platformization to evolve what used to be a simple app meant for on-field usage into a full-fledged, extensible platform that has enabled us to experiment, iterate, and simultaneously deploy variants of home inspection methods seamlessly. Here's how.
Platformization journey of the home assessment process
At Opendoor, we have taken a principled stance: our internal tooling roadmaps and architectures are designed from the ground up to unlock centralization, digitization, self-service, and automation. Platformization is a key enabler to make this all possible. However, making the interior home assessment process digital and self-serve is a paradigm shift in how we have thought about assessments in the past.
In the early stages, we designed assessments emulating how the industry works today, enabling a human operator with a mobile app to conduct in-person assessment. The data was collected and sorted through a rudimentary workflow with multiple other operator roles involved. We didn’t necessarily need to solve for scale in that phase, rather we solved for the quality, granularity, and accuracy of the information that was being captured. As a result, we saw an improvement in the consistency and quality of our assessments.
Soon enough, it was time to upgrade the glorified checklist app and the supporting systems into something much more intelligent. Through our platformization work, we built and automated a post-assessment human review by systematically leveraging data and market insights. As a result, we were able to meaningfully impact the time we take to present the final offer to our sellers and improve the consistency and accuracy of our results, which translated to a reduction in outliers. Essentially, we moved to an escalations-driven system where data elements of captured information that were beyond defined bounds required human intervention. The rest just automatically worked its way through all the transformations.
"Through our platformization work, we built and automated a post-assessment human review by systematically leveraging data and market insights"
In a typical journey once we acquire a home, a series of events take place until the home is resold. Our data models and business logic related to the repairs and renovations phase in the journey were tightly coupled with our monolith. When thinking about building for a 5x scale, extensibility of the tooling and models was critical. We deemed it necessary to decouple, and rewrite, as needed, the relevant systems and services during this evolution. We even built pggen, a database library that enabled us to rapidly generate clients and pull data from three or four databases during the backfill portion of the migration. Pggen can automatically generate structs and infer relationships between tables based on foreign keys. Along with its streaming support, this makes it beneficial for quickly putting together ETL scripts for a backfill.
Foundational system for self-serve
Until now, the concept of assessments was monolithic. It was basically one of our expert operators going to a home to capture information and pass it along as a single assessment report. But when the pandemic hit, it became clear that virtualization was mission-critical.
We enhanced our tooling ecosystem to facilitate customizable assessment templates, which enabled us to make interior and exterior assessments mutually exclusive.
By using a highly configurable visit management and information capturing platform, we were able to evolve it for our business needs over the past couple of years. Built originally for a single-visit and in-person assessment, today it's capable and flexible enough to facilitate emerging use cases. And with this foundational work in place, the team was able to build self-guided home walkthroughs on top of it.
The challenge of this phase was the consolidation of multiple assessment components into a single unified report without breaking any downstream dependencies. Our engineering team tackled this seamlessly and we were able to gradually migrate away from the old and hardwired information capturing system to a newer dynamic, multi-purpose, and highly flexible system. This not only unlocked the strategic rollout of self-guided walkthroughs as discussed previously, but also the capability to mix and match localized versions of assessment templates based on market nuances.
We have been constantly seeking feedback from customers to improve the user experience. In the months ahead, we want to rethink the assessment scheduling process from the ground up in an attempt to remove as many inefficiencies and human interventions as possible.
We are conducting experiments, leveraging machine learning and image processing, to more closely examine potential defects in the home and value the home with minimal human intervention. This will translate to a better customer experience and an improvement in the quality of our assessments allowing for more accurate repair assessments. We think work can be done to reduce dependence on human judgments from the assessment process. And we think we will get there soon as the leader in proptech.
We believe the future of buying and selling a home is on-demand, and leveraging platformization to help us build and create digital experiences is one of the many ways we’ll get there. Our vision is to build a one-stop-shop that allows people to move with the click of a button. Self-guided home walkthroughs bring us a little closer to making that future a reality now.
If this type of work sounds interesting, Opendoor is hiring! Head to Opendoor’s careers page to learn more.