Marketing Spark (The B2B Marketing Podcast)

The episode surfaces a handful of genuinely useful ideas—fraudsters deliberately game whichever KPI you're optimising toward, the 5/10/20% fraud-severity threshold framework, and the reframe of fraud as a business intelligence failure rather than a media-buying problem—but these are interspersed with lengthy origin story, repeated product pitching, and conference-attendance anecdotes that dilute the per-minute idea count considerably.

“the fraudsters know the KPI you're looking for. So what are they doing? Is they're faking that KPI because you're looking at impressions that clicks”

“From the outside, looking in, in a simplistic way, it looks as though fraud is a media buying problem. But if you take a step back, in a sense, it's really a business intelligence problem”

The claim that ad fraud precipitated the 2000 dotcom crash and the disclosure of FraudGPT as a turnkey bot-building tool are genuinely fresh angles not commonly circulated. The insight that optimising toward a KPI actively worsens fraud (because fraudsters target that same signal) is counterintuitive and non-obvious, but the broader 'fraud is under-appreciated and you need a third-party audit' thesis is standard category-education content.

“Fraud was what killed the market back then and back in 2000”

“There's an application called Fraud GPT that will write the bot. And then when you're done writing the bot. You say, okay, how do I deploy this? It will give you step-by-step directions”

Rich Khan is a genuine long-tenure practitioner who built fraud-detection software himself in 2004 when no commercial solution existed and has operated in the space for two decades—well above the average podcast thought-leader. The episode is weakened by the conversation frequently sliding into sales-pitch mode (free audit CTA, trade-show counts, NDA policy), which limits how much hard-earned practitioner wisdom actually surfaces.

“I started my first digital marketing company in 93”

“back in 2004 and five, I figured, let me just write it myself. So I wrote it myself, solved the problem for my clients”

The episode is stronger than average on concrete numbers—$750B global digital-ad spend, $160B fraud estimate, 25–37% average fraud rates across their network, 90%+ fraud on a single university channel, $2.5B identified for clients, and a specific technical tell (all Android phones controlled by the same MacBook Pro). Some headline figures feel loosely estimated rather than sourced, and named company examples are withheld due to NDAs, capping the score.

“we've been watching fraud this year grow across our network. We generally say on average, we find 25 % fraud on when we bring on a new vendor. But last month it was over 37%”

“they had 96, 93%, something north of 90 % fraud”

The host structures the conversation logically and moves through useful topic areas, but he explicitly opens with 'a softball question,' never challenges the 99.999% accuracy guarantee, never pushes back on the dotcom-crash attribution, and allows the guest to repeat the free-audit CTA multiple times unchallenged. Questions are setup prompts rather than genuine probes, and there is no productive disagreement across 34 minutes.

“Why don't we start with a softball question, get your story and an overview of Indura”

“what keeps you at night when it comes to a Neura”