How Sift helped Purse build a trustworthy bitcoin marketplace
150K+
Global users
Millions of dollars in bitcoin saved
OVERVIEW
A new kind of marketplace
San Francisco-based Purse is working to create the world’s largest online marketplace. Their ability to offer the lowest prices around, fast and frictionless commerce, and bitcoin payments means that Purse is actively evolving e-commerce with 5,000-10,000+ unique orders per month. With their two core product features—Name Your Discount and Buy Now—Purse has reached a global audience by matching bitcoin-holders (Shoppers) who wish to purchase goods from Amazon with individuals (Earners) seeking to liquidate their Amazon gift card balances to fulfill orders and receive bitcoin in return.
Purse manages this multi-step transaction through its escrow system to hold bitcoin funds. Since their user base of 150,000+ is truly international, business runs around the clock, with 77% of visitors coming from desktop, and 23% mobile. With a native app launching this Fall, mobile traffic is expected to grow. Purse’s work is a constant, a 24-hour business unconstrained by geography, currency, or time zone, and they process millions in bitcoins every month.
CHALLENGE
Turning virtual currency into real-world economy
In an effort to build a brand around security and trustworthiness, Purse has implemented a guarantee of up to $10,000, which they call the “Purse Guarantee.” If a shopper is ever on the receiving end of fraud, the guarantee is in place to provide peace of mind and ensure that shoppers are protected. This also means that fraud could quickly become very expensive for the small and agile Purse team.
Payment fraud is especially unique at Purse, because with bitcoin, there is no issue with chargebacks; every bitcoin transaction is final and irreversible. Their real challenge is in detecting and removing malicious actors attempting to game their bitcoin escrow system by purchasing items for Purse shoppers with fraudulent/hacked Amazon accounts. Mitigating and preventing fraud is a large part of Amazon’s business—but they can’t catch everything. So, in order to provide their shoppers with added security and the same familiar protections as traditional commerce they’ve come to expect, Purse had to think differently.
As a young startup, Purse initially worked to combat fraud with internal tools, their customer service team digging through internal databases to identify red flags that correlated to bad users. Unfortunately, this practice required three full time support staff committed to fraud management and review, since each member had to investigate 100-150 cases per day, spending hundreds of valuable people-hours every week. With interest in the site only growing, a manual review-only solution was unscalable.
SOLUTION
Maintaining agility with automation
At the recommendation of another company in the cryptocurrency space, Purse decided to look into a machine learning solution to help them speed up the review process and scale with them as they grew. The uniqueness of Purse’s internal order and user management system required close collaboration with the Sift team. Nonetheless, within two weeks, the solution was fully integrated via webhook, allowing Purse to pull Sift’s findings and data points directly into their order management system.
Automating on these findings allows for a more efficient team. For instance, using the Sift Score to auto-ban users over a certain risk threshold gave the Purse Customer Support team the ability to focus on the good customers instead. Since keeping the experience frictionless is key, a non-obtrusive fraud solution is essential to the continuation of Purse’s growth in the U.S. and abroad.
RESULTS
Less time on fraud, more time for customers
Steven and the team at Purse have utilized the many features within Sift to reduce fraud and accurately identify bad users before they impact the site. With Sift Score, network visualizations, and device ID data, Purse is able to more quickly process legitimate transactions. Sift’s machine learning-based solution is constantly improving, which means that Steven’s team trusts that the tool is only becoming more accurate. And the ability for machine learning to digest any and all data allows the customer service to stay small and focused on an excellent, secure customer experience rather than spending hours reviewing transactions that may be good. Fraud is down and business is up, which is great news for this global and growing marketplace.