How to Optimize Campaigns in minutes with the FouAnalytics Domain App Report
The domain app report in FouAnalytics is the quickest and easiest way to make an impact in optimizing your campaigns. The report shows where ads ran (websites and apps) and what types of ad fraud occurred. The worst offenders grid at the top of the page prioritizes the sites and apps draining the most budget so you can take action in minutes.
How to Use the Domain/App Report
The report above classifies impressions under each ad fraud type. When considering which sites or apps to block, focus on sites/apps that have high percentages of fraud and those participating in multiple different types of fraud to make the most impact. Focus on sites delivering the most impressions because that's what is costing the most money.
The rates of fraud you see in FouAnalytics will be higher than what you are used to seeing in other 3rd party verification platforms because FouAnalytics is catching more types of fraud. This is important to keep in mind when you are choosing which sites to block. If you blocked all the sites/apps above 2% in fraud because other platforms commonly report 1-2% ad fraud, then you would be blocking many legitimate sites. Legitimate sites have bots too .
A useful starting point for optimizing in FouAnalytics is to cut sites/apps that have more than 50% fraud. You can block a chunk of the worst offending sites each week so that you can progressively clean the campaigns and visualize the impact in the main dashboard. We recommend you add sites and apps to a block list periodically; then let the data run for a few days so you can see the changes before you make the next change.
When you are reviewing the domain app report grids you can use the check boxes next to the site/app to create a block list within FouAnalytics. Once you have finished going through the report you can simply copy that blocklist into your DSP.
The sites and apps throughout the domain app report are also hyperlinked so that you can look into suspicious sites/apps more closely. Websites are linked to the FouAnalytics PageXray tool which allows you to see details about the number of ad calls and a screenshot of the website. There may be some sites that you would not want to directly load on your computer so this is a safe way to view the site. Apps are linked to the App Store so you can see the details about the developer there.
What Do the Different Types of Fraud Mean?
If you're new to digital media buying, some of the technical language and types of ad fraud might be confusing. The other 3rd party verification platforms never give this much detail as they simply mark something as fraud or not. FouAnalytics was designed not just to tell you a site/app is fraud, but also to give you the supporting details as to why that site/app is fraud.
The different types of fraud in the worst offenders grid can be split into 2 groups based on device or site fraud. Review each different type of fraud below so that you can better understand what the domain/app report is telling you.
Device Fraud
These types of fraud are related to the device where the impression was served. The devices in these types of fraud are not the normal mobile phones, tablets, and laptops that humans use. Fraudsters use a variety of techniques to disguise the device and trick fraud detection so these types of fraud vary based on those different techniques.
No GPU: these impressions are associated with devices that don't have a graphics processing unit (GPU) or a display. This is indicative of a fake browser in a data center because devices that humans use (mobile phones, tablets, laptops) have displays so that you can see the website/app on a screen. In other words, these impressions never had a chance of being seen by a human.
领英推荐
Datacenter: While the No GPU category listed above is indicative of browsers in a data center, the IP address can be a datacenter or not. For example some more advanced bots deliberately bounce their traffic through residential proxies to make it appear they are coming from household IP addresses. That is why you may see some impressions marked as no GPU but also is not from a datacenter. Furthermore, impressions can be classified as a data center based on other information in the device data like operating system or number of cores. For example, the device platform is Linux x86_64, or it has 28-32 cores.
Fake Device: these impressions are associated with spoofed devices. Spoofed devices are designed to trick basic fraud detection by appearing to be legitimate devices. One such example is a mobile emulator which can be used to run 1,000 apps simultaneously and continuously allowing the fraudster to generate impressions 24/7.
Bounced Traffic: these impressions are "bounced" through a residential proxy. This means that the fraudster makes it appear as though the traffic originated from a variety of residential IP addresses. This disguises the traffic so it is not immediately obvious that the traffic comes from a data center.
Site Fraud
These types of ad fraud are based on what happens on the website or app and means that the site is cheating to make more money. If you are used to optimizing with other 3rd party fraud detection platforms, this may be new to you because most of these types of fraudulent activity are not detected by those platforms. This is also why you'll see overall higher rates of fraud measured by FouAnalytics.
Apps Loading Webpages is a category that needs to be looked at closely because it's not always fraud. When you click a link on Facebook it opens a browser within the Facebook app and then that site may serve display ads. This is a legitimate practice and it's useful for Facebook because it keeps the user within the Facebook app and increases their engagement time. The same thing happens in apps like Flipboard, Reddit, Pinterest, etc. News apps may also load their website in their app so if the entities are related (i.e. Fox News app opening the Fox News website) then this is not suspicious. However, this also happens in flashlight, keyboard, and alarm clock apps that have no reason to be loading a browser. The browser in these sketchy apps is hidden so humans do not even see it and aren't aware this is happening. Even though it is not always fraud, FouAnalytics shows this data so you can decide for yourself. Note that the sites being loaded by the app are cheaters because they purchased this shady form of traffic. When you see this in the domain app report, you should block the website rather than the app.
Popunder impressions load in a separate window underneath the active browser. Modern browsers block popunders from loading without user interaction, so these days fraudsters cause popunders to load with malicious code. The code takes any user action (like clicking a play button on a video) as the signal for user intent and loads the popunder. The user is still able to view the video as intended, so they aren't aware anything sketchy has occurred. This is common on piracy and porn sites as a way to launder the traffic.
The next 3 types of site fraud are all about multiplying ad revenue for the fraudster.
Stacked Ads happen when a site loads more than one ad into an ad slot (stacked directly on top of each other). Even if humans visit this site, only the ad at the top of the stack would be seen, so this is cheating.
MFA sites load a ton of ads in comparison to the amount of content on the page and the ads continuously refresh - think slideshows, listicles, and long articles with ads between each paragraph, etc. The sites are often poorly designed with low-quality content (or content stolen from platforms like Reddit) and have high amounts of paid traffic arriving at the website. Often the paid traffic comes from sources like Facebook, Pinterest, or content recommendation platforms like Taboola or Outbrain. The sites are buying this traffic to make money from the crazy amount of ads they stuff on the page. Because many well-known sites are MFA (or have sections of their site that are MFA) this one is a judgment call. A blunt approach to cutting all MFA is not smart . You should first focus on cutting MFA sites that humans do not visit and cutting sites that are cheating in additional ways like ad stacking and pixel stuffing.
Pixel Stuffing these sites load ads into a 1x1 or 0x0 pixel window. This means that even if the ads were served to humans they could not see them because the size is so small. The site still makes money but your ad never had the opportunity to be seen by a human.
For more in-depth reading