How to Use the Domain App Report in FouAnalytics to Review Sites and Apps

For those practitioners who have deployed FouAnalytics in-ad tags in their digital campaigns, the Domain App report tab is where you go to find the top 10 sites and apps that should be reviewed first, by fraud category. The idea is that while there are thousands of bad sites and apps, we only need to focus on the largest by volume. By reviewing those and deciding which to turn off (i.e. add to block list) we can have the most impact on cleaning the campaign.

Let me show you how to use the Domain App report (long screen shot below). When you first go to the tab, you will see the date range and the quantity. It is usually defaulted to the last 7 days. The quantity is the number of ads served in the time period. Below, you will see 8 categories of fraud: high bot, datacenter, fake device, bounced traffic, high PCA, apps loading webpages, stacked ads, and popunders. These are the top 10 sites or apps that you should review first in each fraud category. Note that when you read across each row, you will see a domain and an app name, the count, the %, and the % of total. The first highlighted row below shows an app called "com.hatunnel.plus" being loaded by googleads.g.doubleclick.net (ad serving domain). There are 149,966 impressions in the time period, 67% high bots. The 149,966 impressions is 1% of the 15.5 million impressions in the time period. This gives you a sense of the significance of this one app -- it was eating up 1% of your impressions.

The blue hyperlink of the app name means you can click that to see it in the Google Play Store. Often you will see fake apps have absurd numbers of reviews, all 5-star, or obviously fake developer information (like a gmail address or a privacy policy that is a google doc link). Once you determine there's something wrong with that app, or simply decide you don't want your ads to run on that app, check the checkbox next to the app name. That compiles a list of domains and apps that you can copy and paste to add to your block list. You will also see pagexray links to the right of domains. Any domain that is not an ad-serving domain will have pagexray links; you can click these to see what the page looks like without risking your own device and browser (some domains have malicious code designed to infect your device when you visit, no clicks required).

If you want to see more than the top 10 in each category, click the "See all" link at the lower right of the category. My recommendation is that you focus on the rows that have > 50% in the % column, which means most of the ad impressions match that form of fraud. If the same domain or app comes up in multiple fraud categories, that is yet another reason to turn them off, by checking the checkbox. Jump to the bottom of this article to see an example of the Executive Summary tab where you go to copy and paste the lists of domains and apps you selected to be turned off.

Once you have gone through the categories of fraud above and reviewed the sites and apps for adding to the block list, you can go to the Executive Summary tab and find two text boxes -- Domains and Apps. This is the list of sites and apps you can copy and paste to add to your block lists, or email to your media agency so they can add to the block list for you.

This is where human judgement and intuition come into play. FouAnalytics surfaces the domains and apps you should review and YOU decide whether to add them to the block list or not. There are circumstances where you may decide to not block them even though there are various forms of fraud. For example, even through peopleenespanol.com had high levels of bot traffic and other on-page shenanigans, the advertiser decided to not block the domain because they knew Hispanic audiences they were trying to reach were still visiting the site. They decided to leave the domain on, and eat the cost of the bot traffic. FouAnalytics is analytics, and does not do blocking based on rules or thresholds (e.g. block the site if >5% fraud); FouAnalytics brings up sites and apps for you to review and "decide Fou yourself."

The categories of fraud should be self-explanatory. For the High PCA category, here is the whitepaper about our entropy analysis. https://www.dhirubhai.net/pulse/why-its-hard-bots-avoid-fouanalytics-detection-the-independent/

The Domain App report was built for my own work flow over the years, so I could more quickly identify the high impact bad sites and domains and recommend to clients which to turn off. By turning those off, each week or two, we progressively clean campaigns and make them better. This way, you don't have to study the dark red and light red in the charts. All of that, plus the supporting data grids, are there as additional background in case you want to understand why something was marked as fraudulent, or not fraudulent.

If you have questions, please ask me. I continue to tune the FouAnalytics algorithm myself, after 10 years, and will continue to do so. I am not sure when the last time the DV and IAS algos were updated, nor who at those companies are doing the tuning.

Happy fraud hunting.