SWYM.ai CEO 谈程序化投放的柠檬流量问题与 AI 供应塑形方案

Wednesday, November 11th, 2026 – 1:00 am

SHARE:

A podcast interview with Ravi Patel CEO & Co-Founder

Running a programmatic campaign is like walking into a grocery store with a budget you have to spend before the place closes, and you have no way to tell the lemons from the cherries until you’ve already paid.

Oh, and if you end up buying a lot of lemons, the store automatically assumes you love ’em and keeps stocking more.

That’s a favored metaphor of Ravi Patel, CEO and co-founder of SWYM.ai, a startup that helps ad buyers make smarter decisions before spending their money. (In fact, “SWYM” is short for “Stop Wasting Your Money.”)

But avoiding low-value inventory (spotting lemons) and finding the high-quality stuff (pinpointing the cherries) is both harder than it sounds and harder than it should be.

The main problem is that SSPs and DSPs “talk at each other and not to each other,” Patel says, on this week’s episode of AdExchanger Talks. There’s no feedback loop between them, so all an SSP sees is which bid requests a buyer won, not whether any of them helped that buyer achieve its goals.

On top of that, DSPs throttle traffic to keep cloud costs down, so they only have eyes on a slice of what’s available. Patel estimates that the largest DSPs only see roughly 40% of available supply, while for smaller DSPs it’s more like 20% or 25%.

And all the while, SSPs just keep sending more of the same. “If you bought a lemon yesterday,” Patel says, “you’re getting a lemon today.”

That’s why the filtering has to happen earlier, he says.

SWYM’s approach involves AI-powered supply shaping, which it delivers through containerization, a fancy way of saying its software lives within an SSP’s own systems rather than sitting outside of them.

It might sound like curation, but Patel is quick to draw a distinction. Curation has historically meant packaging an audience, attaching a deal ID and adding a markup. Supply shaping, he says, judges each bid request on its own and only based on whether it’s likely to lead to an outcome.

“I don’t care that user 12345 lives in New York City and loves to snowboard,” Patel says. “I just care whether that potential bid request can lead this client to a sale.”

Also in this episode: The rise and fall of Patel’s previous startup Ad/Fin, how SWYM helped Georgia-Pacific cut its SSP roster by 80% without losing reach, why he thinks AI-generated sites aren’t necessarily bad if real humans are visiting them and whether the open web is really dying. (Hey, they said the same thing about linear, and it’s still here!)

For more articles featuring Ravi Patel, click here.