Matchmaking at Scale
Enhancing the functionality and efficiency of an intelligent real estate referral system
Overview
The Company
A data-led two-sided marketplace, connecting home buyers & sellers to the top Realtors in the US and Canada since 2005.
This company qualifies consumers who are ready to begin their buying or selling process, and then deliver a warm handoff to agents.
The Product
“The Matcher”: the heart of our service, a matchmaking system powered by a machine-learning algorithm unlike any other in the real estate industry.
My Role as Product Designer & Manager, Strategic Ops
Acting as the connective tissue between ops, product, and data, I led the design and execution of improvements to the Matcher’s UI and performance, and related features in the B2B CRM.
This Case Study
We’ll take a high-level look at the most impactful improvements to the Matcher that I implemented over my ten years at the company.
Playing Matchmaker
As the main revenue stream for the company, the quality of matches between home buyers & sellers and real estate agents is paramount.
Core Challenge: Each buyer or seller has their own needs, and each match has unique parameters to consider, such as:
local market quirks
market fluctuations
business partnerships
agent performance & specialties
With nationwide coverage, how can we ensure that each small-town buyer or big-city seller receives the best service?
Iterative Improvements
From 2012 to 2020, I led many phases of improvements to the backend matching UI, implementing new modes of interaction to align operations with AI-driven technology capabilities. This work also involved redesigning workflows, automating tasks, and optimizing processes, resulting in increased scalability of our offerings.
The driving business priorities behind these changes can be captured within two main directives:
Business Directive
Improve Match Quality
Increase our chances of our agents winning the contract, and bolstering agents’ trust in our referrals through a high-quality, consistent experience
Business Directive
Streamline Inefficiencies
Redesign backend UI interactions to reduce labor and stress for associates, keeping our Matching team lean and free to focus on high-level match performance (instead of sweating each match).
Business Directive in Action
Improving Match Quality: Pain Points
Increase our chances of our agents winning the contract, and bolstering agents’ trust in our referrals through a high-quality, consistent experience.
Agent Preferences & Specialties: These text agent notes captured all agent preferences - like whether they only work with buyers or within specific zip codes. Accommodating these requests meant reading through all notes on each match.
No Filters: When presenting agent options for a lead, the system would display all agents in the geographical area, even those with manual notes that would exclude them.
Same Office: Agents can be competitive salespeople, so it’s best to avoid matching agents from the same office to a lead. However, the human matching associate relied on freeform text notes and manually auditing office addresses to know which agents were cohorts.
Hunting for Previous Match Details: When adding more agents to a previously matched lead, the associate had to click out of the matching interface completely to access the office information on each agent already matched to the lead - a time-consuming search just to avoid matching office cohorts.
Matching Interface, before improvements
Business Directive in Action
Improving Match Quality: Solutions
Store Agent Preferences in Standard Formats: Agent preferences are stored in a database readable by AIMS, so only eligible agents are displayed as options for each lead.
Add Filters: Agent designations and attributes are now captured and stored, making them filterable to accommodate business partnerships and lead requests like language or specialty.
Verified Companies Database for Same Office: Introduced a “Verified Companies” process & database to capture, verify, and store shared office cohorts so AIMS can avoid auto-carting them together, and display this data to the Matcher associate.
Display Previous Match Info: Associates can now easily view contextual information on agents previously matched to the lead without leaving the Matching interface.
Verified Companies in Action
AIMS now hides agents in the pool who aren’t a great fit for the lead, as well as agents who are cohorts of the ones already auto-carted.
If the “Blocked” agents are revealed with one of our new filters, we can still add them to the cart, and AIMS will alert us with a red border on the cart’s Agent card if we have two agents from the same office carted on the lead.
Removing one of those cohorts removes the red border alert.
Matching Interface, after improvements
Business Directive in Action
Streamline Inefficiencies: Pain Points
Redesign backend UI interactions to increase efficiency and reduce labor and stress for associates, keeping our Matching team lean and free to focus on high-level match performance.
Tedious Navigation: The original MVP matching interface did not consider repetitive motions and how component placement might impact the associates’ efficiency over time and volume:
1a: Action buttons to add/remove agents from the lead are placed across the screen from the most used component, the “Match Agents” CTA button.
1b: If the lead is in a densely populated area, the number of eligible agents displayed could reach into the hundreds. The associate must manually scroll down and up the page to review agents and view the details of carted agents.
Not Built for Human Eyes: The interface lacks visual hierarchy and there’s little consideration of how humans best capture and process large amounts of data:
2a: All typefaces have similar weights, limiting their scannability.
2b: Agent performance (and assumed fit for the lead) is represented by hard numbers, with little differentiation and labeling, so it takes more cognitive power to make comparative assessments.
2c: Color choices (like the yellow background) cause eye fatigue, and buttons are a missed opportunity to create a visual hierarchy for CTA functions
Matching Interface, before improvements
Business Directive in Action
Streamline Inefficiencies: Solutions
Intuitive Navigation:
1a: Action buttons to add/remove agents from the lead are grouped near the most used “Match Agents” button, minimizing the back-and-forth movements across the screen. The Add/Remove choice is also added to the agent cards in the Cart.
1b: Agent cards in the Cart are now hyperlinked to the corresponding agent card in the pool - with one click in the cart, the pool view scrolls to the details agent card. Also added a floating “Return to Top” button to eliminate excessive manual scrolling.
Scannable by Human Eyes:
2a: Introduced more visual hierarchy to text by varying the weight and placement of key pieces of info
2b: With the support of the Data Science team, we adjusted the output of AIMS so that a “Match Fit” percentage is the primary stat when considering an agent’s fit. This relieves the human from weighing complex data and decisions that AIMS has already considered, making agent comparisons easier.
2c: Interface background colors are now much kinder on the eyes, and accent colors work with data visualizations like charts and graphs to intentionally communicate data and guide user behaviors.
Matching Interface, after improvements
Outcomes
Growth Unlocked
When I first took the Matching helm, leads were painstakingly matched by hand. I knew each market and recognized our top closing agents on sight. I recall days where I matched only 15 leads in a day, and that was a success!
After years of iterative work aligning and improving the UI, UX, algorithm, and operations processes, we scaled our volume at breakneck speed. In 2018 alone, the company processed around 15,000 leads per month, and 250,000 matches in the year.
Fully Automated Matching
The Matcher was always more than just an interface, but over time we were able to translate decision-making from a manual human process (avg. 5 min to match a lead) to full automation (a matter of seconds).
This reduced errors and the need to expand the Matching team to service higher lead volume. My team’s efforts were freed up to supervise and hone high-level network performance and product direction, increasing the value of our contributions to the business.
Groundwork for Partnerships
The company cultivated relationships with some of the most respected real estate organizations in the industry, such as CRS and Berkshire Hathaway.
We were able to leverage past AIMS data integration projects to accommodate key partnership requirements such as automated lead flow throttling to partner agents.
Enabling International Expansion
In 2019, the company expanded their agent network to Canada. As the Matching product owner/designer, I implemented critical product updates to AIMS and backstage structures to manage the intricacies of our new market.
Because of our past work to enhance matching efficiency and capabilities, we were able to accommodate new international requirements easily.