How a major fintech platform stopped viral fraud before it reached customers

Learn how a leading peer-to-peer financial services company identifies emerging fraudcampaigns before they spread across video-first platforms.

• 3.5M+ pieces of content reviewed in first 3 months
• 2.4K+ pieces of fraudulent content caught in first 3 months
• 22 reporting hours saved per incident

”What stood out immediately was how well this fit the way we think about fraud. Wegained visibility into platforms we simply couldn’t monitor before, and now, if ascam shows up it’s automatically flagged for us so we’re able to intervene beforeour user community is impacted.”

Challenge

Fighting fraud on video-first platforms without scalable coverage

For a peer-to-peer finance platform built to expand access to financial services, trust iseverything. But many of this company’s millions of customers have historically lacked access totraditional banking, making them prime targets for scammers.

To mitigate this risk, the company’s security team had already invested in monitoring tools thatprovided coverage across text- and image-based social platforms. But as the platform’s usersincreasingly favored video-first channels, the team faced a growing coverage gap.

Because existing solutions relied on keyword searches and legacy ML models, they couldn't analyze transcripts or video content. And without extensive monitoring, scams, misinformation, and complex fraud operations could easily spread through videos the team might not see—on the very platforms that mattered most.

To close the gap, the security team attempted sweeping video-based social platforms for harmful content. But with millions of legitimate videos posted daily, including a flood of payment requests from content creators, catching the small percentage of scam content in real time wasn’t possible.

And the longer threats went undetected, the more customers were exposed. Once a scam was discovered, the team still had no reliable way to see how far it had progressed, if the same fraud campaign had crossed multiple platforms, or whether the activity was driven by bots.

If scams were successful, preparing incident reports for leadership involved hours of manual work. Teams had to hunt down every mention across tools and assemble accurate summaries with limited information.

The team’s resource-intensive processes were eventually put to the test after a viral video promoted a “glitch” in the platform’s system. The scam encouraged customers to send money to strangers with the promise of getting more in return. By the time the team exposed the scheme 48 hours later, approximately $2 million had already been lost to fraud.

In the aftermath, the team immediately began evaluating video-first fraud detection solutions. At first, they considered an in-house build. But even with an engineering team in place, maintaining and evolving an internal system would require extensive resources.

Instead, the company wanted a partner who could automatically deliver full visibility into video and audio content without extensive manual review. Based on their experiences with previous vendors, they also knew a self-service model was essential—one that wouldn’t require pulling in slow external support for every adjustment.

When the company launched an RFP, Outtake stood out as the only platform truly built to handle video-first fraud at scale.

”Video-first socials were a blind spot unfortunately. We were picking up tens of thousands of mentions every day, but our tools couldn't touch video content, so scams could go unchecked for days before being identified and addressed, a real risk for our operations.”

Solution

Detection that mirrors how analysts think without needing to watch every video manually

The partnership launched with a default deployment, giving the company immediate visibility into potential video and audio threats. From there, Outtake’s team tailored the solution to the company’s unique needs and workflows during a four-week pilot.

During this phase, the team reviewed sample alerts, shared how they would handle cases, and gave feedback on prioritization and response. This input was used to ensure decisions reflected the team’s existing risk framework and how their analysts already operated. By the end of the pilot, the team was able to apply their starpower across a much larger scale of identified fraudulent activity.

Rather than taking on costly internal build, they sidestepped an enormous, unnecessary engineering burden. Unlike a homegrown solution, there are no keyword libraries to maintain, no ML models to retrain, and no infrastructure to manage. Outtake’s system continuously improves through analyst feedback, refining its logic with each iteration so that manual overhead reduces dramatically over time.

With scalable coverage across video-first platforms, the team no longer has to rely on manual sweeps to find the needle of scam content hidden in massive haystacks of social posts. Instead of combing through noise, analysts now start each day with a focused stream of high-signal investigation summaries.

Even better, the platform is fully self-serve. The team can now manage workflows, adjust priorities, and refine detection logic as their needs change. Instead of waiting days for vendor support to make simple changes, the team can finally move at the speed of the threat landscape.

But the real shift is deeper than speed. The security team no longer has to anticipate every possible scam in advance. Emerging fraud patterns are flagged early and brought to their attention quickly, even when they don’t resemble anything seen before. Outtake's platform mirrors how the team thinks about fraud, so emerging schemes are automatically identified in minutes and surfaced via Slack and email—even when threats don’t match any existing rule or known pattern. The kind of 48-hour blind spot that once led to a $2 million loss has effectively been eliminated.

Since Outtake is also well positioned to proactively take down fraudulent assets while monitoring for scams and security risks, the system can also be configured to identify the profile behind attacks and flag malicious links embedded in the content's description. Shutting down these buried threats protects customers from coordinated campaigns that could otherwise fly under the radar.

Incident documentation used to be a time sink. Compiling reports once required hours of manual work across multiple tools, but now analysts can generate natural language reports with analytics in a single click.

These reports show details that weren’t available before, like how far a scam has spread, whether it has crossed platforms, and whether it came from bots or platform users. Instead of scrambling to reconstruct what happened, analysts walk into meetings with the full story already assembled—and proof that threats were tracked and documented in a timely manner to mobilize response.

”What mattered most was consistency and control. Now we can apply the same judgment of our best analysts across far more activity than before with Outtake’s platform. When new patterns emerge we also identify and adapt responses faster than we could before.”

Results

2.4K+ pieces of fraudulent content caught in the first 3 months

The team didn’t just improve threat detection. They gained the clarity needed to act decisively. Investigations are more focused, and leadership has greater confidence in how fraud risk is being managed. The impact is clear:

• 3.5M+ pieces of content reviewed in first 3 months
• 2.4K+ pieces of fraudulent content caught in first 3 months
• 22 reporting hours saved per incident

“Outtake gave us full visibility into every platform our customers are using and theproof to show leadership we're catching threats before they ever have the chance tospread.”