Commercial Real Estate Investor Search
First-generation product built on a sophisticated machine-learning data platform.
Ten-X is a company that sells real estate online. The division where I work is focused on commercial real estate, an industry that is largely controlled by brokers. Accurate information about commercial real estate properties, sellers, and buyers is hard to come by, as most brokers tend to hold their cards close to their chest.
Ten-X set out to build their own data source (codenamed "Krang") for use in machine learning applications. The first generation of those applications was my first project at Ten-X, a tool used internally to generate lists of potential buyers for specific properties.
The Vision
This forward-looking mockup, created for a board deck, was designed to deliver the sizzle of machine-learning applications. The steak (i.e., the predictive features on display here) is still several years away, but this was effective at generating enthusiasm (read: funding) for the project.
The first real step, of course, was to come up with an MVP.
User Research
It had been decided before I arrived that the first application we built would support the inside sales team. A nice, managably-sized group of users, they still had a broad range of needs that would require some research before a suitable MVP could be chosen.
Ten-X had a talented group of user research people that had already been conducting interviews with key members of the team. Based on the interviews they had recorded (and a few that I conducted myself), I was able to put together a diagram of a typical pre-sales workflow.
This revealed a clear pain point: lead lists (which can have up to three thousand entries) were assembled by a small, over-worked research team...a process which required hours of manual data wrangling.
Minimum Viable Product Proposal
The team needed a tool that would search through multiple data sources at once and deliver a single list, ready for consumption by the CRM system that was already in use. The diagram below illustrates what the lead-generation workflow might look like with the new tool in place.
Filter Library
The goal of the tool would be to narrow down a huge database of potential buyers into a shortlist that could be followed up on by the sales team. The primary interface was a series of filters, i.e. a way to enter information about the property for sale, in order to generate a list of potential buyers for that property.
I started putting together a library of different filters we would need, each with different interface options.
Wireframes
The inital delivered version of the Investor Search tool consisted of two templates:
- The Search Page, which included the filters and an easy way to export the final list.
- An Investor Detail Page, where additional information about each investor is displayed, to help the sales team make better connections.
Learnings
- Initial usability tests with the lead-generation team were very successful.
- The ability for the product to successfully identify buyers for specific properties improved over time, as data scientists trained the algorithm. In lead lists of 300 people, the tool will find the winning buyer 70% of the time (tested using historical data).