How KSL improved manual review efficiency while reducing content fraud
5x
Increase in manual review efficiency
80%
Reduction in time spent on fraud management
OVERVIEW
Creating a local online community
KSL started off as a television company within Deseret Management Corporation, gaining affiliation with NBC in the 1990s. As part of the overall brand, KSL Classifieds was built in the late 1990s to offer Utah residents a local online classifieds platform to complement the community news product. In 2009, Deseret Digital Media was formed to separate digital properties from traditional radio and TV media properties. As part of this move, the DDM Marketplace business unit (within which KSL Marketplace resides) was created to further distinguish local commerce marketplace entities from the online media entities. So far, the move from traditional news agency to online and digital platform has been nothing short of successful: Utah is the only online classifieds market in the country not dominated by Craigslist, thanks to KSL’s Marketplace.
The Marketplace currently sees over 235 million page views per month and is growing rapidly as KSL aspires to become a national brand. Unfortunately, with all of the success and attention, KSL’s classifieds platform also became a popular target for fraudsters. In response, they hired Eric Bright as Vice President of E-commerce and charged him with not only growing revenue, but also stamping out KSL’s fraud challenges.
CHALLENGE
Separating trusted users from the suspicious
KSL.com, like any classifieds marketplace, is a user-driven platform of both buyers and sellers, making trust a key ingredient to success. And with a growing percentage of fraudulent postings, KSL was suffering from an existential problem. Bad users were scamming legitimate users from all sides: publishing fake listings, taking over legitimate customer accounts, and running scams from hijacked accounts. Malicious users were also harassing the sellers of real listings, trying to scam them out of their goods and services.
The main challenge Eric faced was not only finding and eliminating existing fraud, but also blocking suspicious users as they tried to re-access the site after one device or account was banned. KSL needed the ability to auto-ban bad users and repeat offenders. Fighting an imposing fraud rate of 75-80% in some of the more popular sections of the site, KSL’s sole fraud analyst wasn’t able to keep up with the demands placed on their internal fraud tools and manual review process—so the team brought in a traditional fraud management vendor.
After two years of struggle, that fraud solution still wasn’t fully in place. KSL’s fraud analyst had to review every order to train the system and the solution was slow to integrate. After finally getting the product online, KSL discovered that not only was this solution inaccurate and ludicrously expensive, but it also wasn’t scaling. Instead of adapting to KSL’s needs, the vendor recommended the Marketplace team hire five more fraud analysts to overcome the solution’s deficiencies. After this painful experience, KSL was ready for a powerful and accurate solution that could drive automation and reduce (not increase) their investment in manual review.
SOLUTION
Accuracy and automation
Hiring five additional people just to review fraud is expensive, so Eric started evaluating Sift’s capabilities as an alternative. He found that trialing the Sift solution was easy; the full integration took just six weeks and they started to see results immediately. With just a few weeks of labeling, the KSL team trained their customized model to be so accurate that they could confidently rely on Sift Score ranges to automatically approve, reject, and review transactions. Sift’s accuracy in pinpointing fraudsters allowed KSL to identify and shut down fraud faster than ever before. That meant the KSL team could focus their energies on the truly suspicious users, keeping good customers happy with quick approvals
RESULTS
Visualizing fraud to find it fast
Before Sift, KSL’s average manual review time was 5x longer than it is today. By moving to Sift, KSL estimates their total “cost to sort”—the man-hours required to review suspicious orders – has dropped nearly 80%. The flexibility and agility of the Sift solution allows KSL to quickly and efficiently auto-reject users and any connected fraudulent accounts. Eric’s team relies heavily on Sift’s Network Visualization, Connected Users, and Formulas capabilities, which give them the ability to easily and automatically find who to ban from their site. In one instance the KSL team caught a single user who operated thousands of fraudulent accounts, all because Sift linked the “unique” accounts by unveiling shared attributes and intuitively displaying this fraud network using advanced data visualization techniques. Now, their fraud team is proactive and agile. Through Sift’s quick and intuitive interface, KSL has been able to distribute fraud monitoring responsibilities across multiple existing team members to provide near round the clock fraud coverage without adding headcount.
Sift is now integrated with and protecting all of KSL’s sites, working in real time to detect malicious users. Where these sites were once bogged down with spammers and scammers, their loyal user base has noticed a shift in the numbers; fraud is down 33%-54%—depending on the site—and the community can once again enjoy this platform for real exchange and conversation. KSL trusts Sift so much that they use Sift Scores to auto-act on orders, processing hundreds more per day and allowing Eric and his team to focus on growth, not fraud.
