Bad Traffic Brought Visitors to Our DMEXCO Booth, but Good Traffic Kept Them There

The CEO of an SSP stopped at our stand with a fraud problem. Most of his traffic comes from APAC, where the invalid traffic share is worst. He came to talk about residential proxies and device counts. Then the conversation moved to Places, and to what his clean traffic could be sold for if a buyer knew where it was. He left having booked a call with his CTO.

Over two days at DMEXCO, I had versions of that conversation again and again. Someone walked up worried about bad traffic, then realized the same IP data can make their good traffic worth more.

Below is what publishers, SSPs, DSPs, agencies, and telcos discussed with me and my colleagues.

How Much of This Traffic Is Real?

Publishers described bots and AI crawlers spiking their traffic and slowing their sites. An ad ops architect at an Italian publisher sees a million bids a day and does not know how much of it is human. A yield manager at another said invalid traffic was hurting both their reputation and their advertisers' return. SSPs wanted to filter bad publishers before the auction. A CTV network wanted to run its own scans alongside its existing vendor.

What connected these conversations was the wish to see the underlying signal. Verification tools return a verdict. The people I spoke to wanted the data behind it, so they could set their own thresholds and explain a decision to a customer. A DSP founder put it as wanting lower-level signals to make more precise decisions.

Residential proxy detection answered this. Once someone sees that a slice of their "real user" traffic is routed through residential IPs that belong to proxy networks, the rest of the conversation changes.

The Same Data Makes Good Traffic Worth More

This is where Places came in, and it came in most often with people who had started on invalid traffic. An IP tells you if traffic is not bot-related, and it can also tell you the traffic is in a retail park, a conference center, or an airport.

The requests were specific. The ad unit of a German telecom cannot segment its traffic by location at all and wants to sell publishers audiences like "customers in Berlin shopping for trainers." A property portal wants to show ads for new developments to people near the site. A broadcaster wants to enrich impressions before they reach the SSP so they clear at a higher bid. A data company that works with banks wants every IP in a given postcode. Agencies asked about footfall, venue targeting, and radio listeners on a commute.

A martech platform lead called segmentation that does not depend on cookies "a win." Two large data companies, an agency holding group and a DMP, want IP as a layer that survives cookie loss.

A Common Concern: "IP Data Is Not Accurate"

More than one conversation opened with this opinion as a settled view. One telecom team arrived convinced of it. We walked them through how we build and verify the data, and they booked a call for the following week.

A publisher's programmatic lead uses MaxMind, has gaps, and needs reliable resolution to enrich their warehouse. A CTV platform in Ukraine also uses MaxMind and asked us to put in writing where our geolocation differs, so he could take it to his CTO. An ad server CEO brought his own customer to the booth because they had hit accuracy problems with their current data and could not use its API. Both are set to start an evaluation in October.

In every case the objection came from experience with a single provider, most often MaxMind. What changed minds at the booth was showing how the data is built: where the location signal comes from, how we verify it, and how often it is refreshed. Once people saw that the method differed, the conversation moved from whether IP geolocation works to how soon they could test it.

What We Took Home

Most people came to us with one question: how much of my traffic is real? The ones who got the most from the conversation left with a second one: what do I know about the traffic that is?

An IP address answers both. It tells you whether traffic is a bot or a proxy, and Places tells you the kind of venue the traffic is in, whether that is an airport, a university, a retail park, or a hospital, which is enough to build an audience without a cookie or a consent string.

If either question sounds like yours, we would like to hear it. Talk to the team.