At Smadex, we know that protecting campaigns starts with clarity, smart detection, and strong technology. That’s why we sat down with Oriol Malvehy, our Senior Product Manager, to talk about today’s most common fraud schemes, how Smadex blocks them in real time, and why post-install signals are key to keeping performance clean and reliable.

“Ad fraud is prevalent in digital advertising, where malicious activities manipulate ad metrics to deceive advertisers. This fraudulent activity can take many forms, from bots generating fake impressions and clicks to sophisticated schemes that steal credit for real user actions. The primary goal of ad fraud is to move ad spend from legitimate advertisers to fraudulent parties, leading to wasted budgets and inaccurate data that can skew campaign optimizations.

The biggest scam nowadays is fake installs. This is when a fake download of an app happens without a real person behind it. Think of it like a robot or a bunch of fake phones “installing” your app repeatedly. These fake installs are designed to look real just long enough to get paid, but they’re completely worthless. Since no actual person is using the app, they never sign up, never make a purchase, and never do anything else you’d want a real user to do. Ultimately, you’re paying for a ghost download that brings no value to your business.”

“Smadex relies on a multilayered system to protect against ad fraud. The first layer is all about speed and works in real time when an ad request reaches our servers. This allows us to run quick checks and immediately block suspicious traffic before an impression is even served. It works as a first line of defense that filters out the most obvious cases of fraud at the very entry point.

The extra layers add a deeper level of protection, both at the campaign and global scale. At the campaign level, we constantly monitor key performance indicators, like IPM and post-install metrics, to quickly flag sources that don’t behave as expected and cut them off before they drain the budget. At the global level, we analyze data across all campaigns to find recurring fraud patterns, permanently block publishers that consistently show fraudulent behavior, and clean our datasets. That last step is crucial because it keeps the information feeding our models accurate, which makes our optimization and decision-making much more reliable in the long run.”

“Real-time fraud detection is critical because of the immense volume of traffic being processed every hour. This high volume allows for sudden and massive spikes in fraudulent installs, which can quickly drain a campaign’s budget. Detecting fraud instantly prevents these spikes from causing significant financial damage. By catching and blocking fraudulent activity as it happens, you ensure that ad spend is not wasted and that your campaign’s performance data remains clean and accurate.”

“Based on current trends, the most common ad fraud today is driven by smaller publishers and malicious sources that focus on generating fake users and fake installs. These fraudsters create a superficial sense of success with a “too good to be true” Impression-to-Install rates. This tricks advertisers into thinking their ads are highly effective and drives up the cost of traffic. The real giveaway, however, is what happens after the install. Since these “users” are not real people but bots, they never perform any genuine post-install activity like making a purchase, signing up for an account, or simply engaging with the app.”

“Fraud is a sophisticated and ever-evolving threat. The best way to strengthen ad fraud protection is to go beyond install-level data and focus on post-install signals. It’s also key to work with partners that use accurate and effective fraud detection systems.

While a fraudulent install might look legitimate on the surface, a fake user will never take meaningful actions like making a purchase, signing up for a newsletter, or reaching a key level in a game. You can build a comprehensive user profile by analyzing multiple post-install signals and effectively distinguish between genuine and fraudulent behavior. Smadex, for example, looks at close to 10 extra signals, allowing for a much more nuanced and accurate decision on whether the traffic we see from publishers is real or not.”