Ad Fraud: Project Shield Saves $180K in 2026

Ad fraud is still a huge budget drain on digital campaigns, and fighting it requires way more than just basic blocking. By 2026, the schemes are so complex that you need a proactive, data-fueled defense that can predict where fraud will pop up next, not just react to it. Smart marketers are protecting their investments by getting ahead of the problem with predictive intelligence.

Our “Project Shield” campaign gives a clear picture of the fight against persistent ad fraud. We were targeting B2B SaaS decision-makers in North America with a $180,000 budget over a six-week duration, all focused on generating leads for a new enterprise cloud migration service. The main goals were a Cost Per Lead (CPL) under $150 and a Return On Ad Spend (ROAS) above 2:1. Going in, we projected about 15 million impressions, a Click-Through Rate (CTR) of 0.8%, and 800 qualified leads.

We ran a multi-channel strategy using programmatic display through a major DSP, LinkedIn Sponsored Content, and some targeted native advertising. All the creative assets were built around problem/solution stories that focused on data security, scalability, and cost savings. Our targeting was tight, layering on firmographic data, job titles, and intent signals. For programmatic, for example, we zeroed in on domains popular with IT Directors and CIOs and cross-referenced them with IP reputation scores from our ad verification partner, Anura. Even LinkedIn, which you’d think is safer, had its own issues with fake profiles and engagement bots, which forced us to get really strict with audience exclusions and watch on-site metrics like time on page like a hawk.

The early results were a disaster. In the first two weeks, we saw a massive number of reported impressions, over 7 million, but our CTR was stuck at 0.15%, a fraction of our 0.8% target. Even worse, the Cost Per Click (CPC) looked great on paper at around $0.50, but the conversion rate from those clicks to an actual qualified lead was less than 0.05%. That was an immediate red flag. We were getting tons of “clicks” that produced zero meaningful engagement. Our CPL was north of $1,000, which meant the budget was just evaporating. The skewed data was the real poison here, making it impossible to tell what was working or how to optimize. I’ve seen too many campaigns go off the rails because the team was afraid to admit how much fraud was in their traffic.

Our first move was to tear into the traffic quality using our chosen digital security tools. We integrated Integral Ad Science (IAS), a real-time ad verification platform, directly into our DSP, which let us see impression-level data and immediately flag suspicious IP addresses, data centers, and clear non-human traffic patterns. At the same time, we used ForensiQ’s behavioral analysis to spot weird activity like impossibly fast clicks, jerky mouse movements, and user journeys that made no sense. With global ad fraud losses projected to hit $100 billion by 2023 according to Statista, the scale of the problem we were seeing wasn’t a surprise.

IAS confirmed our worst fears: about 60% of our programmatic impressions and a staggering 75% of the clicks during that initial phase were flagged as fraudulent. That was the data point we needed. We immediately implemented aggressive pre-bid filtering using the live data feeds from IAS, which meant we could exclude entire IP ranges, flagged domains, and any publishers that had a history of low viewability and high invalid traffic (IVT) rates. We also had to adjust our bidding strategy to prioritize inventory with better quality scores, even though it meant paying a slightly higher CPC for clicks that were actually real.

On the LinkedIn side, the fraud looked different. It wasn’t so much bot traffic as it was fake engagement from click farms or empty profiles liking our posts. To combat this, we tightened our audience filters, getting much stricter about profile completeness, seniority levels, and recent activity. We also had to manually review engagement spikes on our sponsored content, looking for patterns of generic-looking profiles with no relevant connections liking or sharing our posts. It meant we had to do some manual, qualitative checks, but that human oversight was just as important as the automated tools.

Mid-campaign, we made a few big moves. We pulled 30% of the programmatic budget out of the open exchanges and put it directly into private marketplaces (PMPs) with publishers we had direct relationships with and trusted for high-quality audiences. We also jacked up our minimum bid floor on programmatic campaigns to help price out the low-quality, fraudulent impressions. Finally, we kept our exclusion lists updated daily based on the latest IVT reports. That constant cycle, identify, exclude, re-evaluate, was the core of our whole fraud prevention strategy.

The results from these changes were huge. By the time the six-week campaign ended, our numbers had completely turned around. Total impressions fell to 9.5 million (a 37% drop from what we first projected, but that was just us cutting out the fraud), but our CTR shot up to 0.72%, right near our target. The quality of the clicks improved so much that our click-to-lead conversion rate hit 1.8%. Our final CPL came in at $135, well under our $150 goal, and the ROAS from closed deals tied to the campaign hit 2.8:1, beating our 2:1 target. It just proved that by focusing on genuine human engagement, we got far better business results, even with fewer total impressions. A lower volume of high-quality impressions will always beat a high volume of junk, people get so caught up in chasing scale that they forget this.

Even with all our filtering, the bots kept evolving. That was a persistent headache. We started seeing new fraud patterns, like bots that were getting better at mimicking human scrolling or cycling through IP addresses more rapidly to avoid detection. This just means you have to be constantly vigilant and keep adapting your rules. We also learned to cross-reference different data points. For instance, if a publisher was reporting high viewability but our own analytics showed suspiciously short session durations from their traffic, that was a red flag that needed more digging. The fraudsters are always adapting, so anyone claiming they’ve “solved” ad fraud is either lying or not paying close enough attention.

Our post-campaign analysis just confirmed what we already knew: you need a layered digital security approach. The combination of pre-bid filtering, real-time verification, and constantly optimizing based on granular data works way better than any single-point solution. The initial investment in good fraud detection tools easily paid for itself just by stopping the budget bleed, which in turn boosted campaign efficiency. We ended up with a $135 final cost per conversion, a massive improvement from where we started. Without these fixes, the entire campaign would have been a write-off, not because the strategy was bad, but because the data was garbage.

This whole experience also made us push our ad tech vendors for more transparency. We started demanding more granular reporting on IVT rates down to the publisher and placement level, something not all platforms want to provide. We need that level of detail to make smart decisions about where to put our money in the future. The IAB is always talking about transparency and standardized metrics for a reason, it’s one of the few ways to fight fraud effectively as an industry.

On the creative side, we noticed that using interactive elements or longer-form content on our landing pages seemed to deter simple bots, which tend to bail on complex interactions. Our video ads got good engagement, but we had to be really careful with video completion rates. Some bots are programmed to “watch” a video for a set amount of time just to appear human, so we had to look at what happened *after* the view, like subsequent page visits or form fills, to know if it was a real person.

Project Shield was proof that even though ad fraud is everywhere, you can manage it if you have the right tools, strategy, and stay vigilant. That $45,000 we saved by not paying for fraudulent impressions and clicks was redirected to legitimate inventory, and it directly helped us beat our ROAS target. Effective fraud prevention directly maximizes the return on every dollar you spend on reaching real people.

Because the threat of ad fraud isn’t going away, marketers have to build fraud prevention into their strategy from day one. It has to be an integral, ongoing part of every digital advertising campaign, requiring constant adaptation and the use of advanced verification technologies to stay ahead of new threats.