Removing invalid syndicated traffic cut cost per signup by 18%

This started with something small. 

Sales teams were struggling to reach new leads. Calls weren’t connecting. Emails weren’t getting replies. Conversations that should have been happening weren’t. 

On the marketing side, nothing looked broken. 

The account was spending six figures per month on Microsoft Ads and scaling. Lead volume was rising. Cost per signup was stable. Platform reporting suggested performance was steady. 

When lead volume increases but sales can’t reach people, something is wrong. 

The question wasn’t whether leads were being generated. They clearly were. The question was whether they were real.

Nothing in-platform looked obviously broken. Forms were completed. Contact details appeared legitimate. There were no spikes in junk submissions. 

The marketing funnel was full but the sales pipeline wasn’t moving. 

We decided to segment performance by network and placement and found that the unreachable leads traced back to Syndicated Partner inventory. 

We suspected some of that traffic was fake.  

AI-driven bot traffic has become harder to detect. It can mimic browsing behaviour, complete forms convincingly, and pass validation logic. At campaign level, this traffic blends in. Only when examined by source does the distortion become visible. 

The working hypothesis was simple: part of the Syndicated Partner traffic was generating convincing but non-reachable leads. 

The intervention

Rather than exclude broadly, we ran a controlled audit. 

Referrer tracking was added directly to the lead forms. This created an independent record of where each signup originated. 

High-volume domains were reviewed against downstream contact rates. Were these sources producing leads sales could actually reach? 

A small group of repeat offenders emerged. They were driving consistent signup volume with almost no successful contact.

Why the system didn’t correct itself 

Automated bidding systems optimise toward the signal they are given. Here, that signal was a completed form. If a placement generates conversions at target CPA, the algorithm allocates more impressions to it. 

For partner sites, more impressions mean more ad revenue. If invalid traffic produces convincing conversion signals, the loop sustains itself. 

Once confirmed, those sources were excluded. Budget was reallocated away from the associated Syndicated Partner placements, with total spend remaining flat. 

What changed 

The shift was immediate. 

• Spend: Flat  
• Cost per signup: ↓ 18%  
• Lead quality: Improved materially  
• Funnel reporting: Trust restored 

Sales teams could reach prospects again. Reported performance began to align with actual outcomes. 

Bots are no longer obvious. They can generate realistic behaviour and pass surface-level checks. 

Stricter form questions or IP filters are often not enough. The distortion only becomes visible when traffic is examined at partner level and compared to real-world outcomes. 

In this case, removing invalid Syndicated Partner traffic reduced cost per signup by 18%, and the numbers started to mean something again. 

Where do you want to take this next?

If you suspect something similar is happening in your account, you can explore more case studies, learn how we approach audits, or book a strategy session to review your setup.