Real-estate startup Cozy is streamlining how renters and property managers connect. With over 400,000 users on the platform, Cozy processes nearly $1 billion in payments every year.
A marketplace case study
How Cozy creates a safe platform for renters and landlords
In fraud prevented
constantly falling fraud rate
Automation on accurate scores
Told from the perspective ofKevin Collins,Head of Trust & Safety
“Sift Science has been extremely helpful. I would be lost without it.”
Making renting easy for everyone
Cozy is a Portland, Oregon-based startup that offers a platform to streamline interactions between property managers and renters. Using this consolidated system, renters can connect with landlords, apply for properties, and pay their rents through the simple web interface, accessible on desktop or mobile.
On the landlord side, Cozy makes recurring payments and property management simple and automatic. Additionally, the platform allows property managers to market their rentals and screen their prospective tenants online. With a commitment of making renting easy for everyone involved, Cozy processes close to $1 billion in rental payments every year. The company boasts 350,000+ property managers, landlords, and tenants in cities throughout the United States.
Maintaining trust in the platform
Renting an apartment is a nearly universal experience, and Cozy’s user base has only grown as tech-savvy young professionals have begun their hunt for a place to live. With listings and renters in 13,000 cities nationwide, fraud hasn’t been far behind. As a combination of a marketplace as well as a payment gateway/ processor, Cozy experiences both payment fraud as well as content abuse in the form of fake rental listings.
Payment options on Cozy are only growing – expanding beyond ACH to credit card – and wily fraudsters have begun using fake rentals to ask for wire deposits from unsuspecting renters. In order to stay ahead of the issue and prevent loss of trust in the platform on both the renter and landlord sides, Cozy’s fraud team of 1 needed a solution that could keep up with their wide user base and prevent fraud before it resulted in losses.
“When it comes to seeking and destroying fraud, it’s been just me analyzing the data since our first fraud loss in July 2014 until April 2017, when we added an analyst.”
Fighting fraud with a team of 1
As a lean and agile startup, Cozy brought on Kevin Collins and leveraged his experience in tracking fraud with online communities. Having sourced solutions and weighed his options, Kevin pushed for a machine learning solution that could scale with the business and be predictive in its modeling. With just a small team dedicated to fraud, Cozy couldn’t afford to be reactionary.
After trying a few different solutions that didn’t work for Cozy, Kevin found Sift Science. He was able to demonstrate Sift Science’s immediate value to the company, and a quick integration followed.
“We had everything up and running within a few days of beginning our work, brought on by a 2-person engineering team only tapped for a few tweaks every few weeks. Once we integrated, the model just clicked. It worked right off the bat and has only gotten more accurate. Even better, it’s been easy for us to manage.”
Business is booming, fraud is fizzling out
Now, although business has grown, fraud rates have remained low. Kevin is efficient in his fraud management, automating based on Sift Score. When further investigation is required, he can do so effortlessly with the Sift Science Console, where he explores connected accounts and account activity.
He has a single stop to train Cozy’s machine learning models, manage the company’s user base, and better understand their platform’s traffic. Best of all: since integrating, the Sift Science platform has accurately identified over $1 million in fraudulent payments, allowing Kevin to stop these attempts before they could result in costly chargebacks – and Cozy’s fraud rate is continually decreasing.
“We’ve seen excellent business growth, while our fraud rate has continued to decrease. We lost under $10,000 in 2016, which is pretty incredible given how much we process.”