FouAnalytics is not just about fraud detection, it's about better digital marketing

Since I've been researching and writing about ad fraud for a decade, it's understandable that folks think of FouAnalytics as fraud detection. It does fraud detection, but marketers and I use it for more than just that. I called it "analytics" because it give us more detailed data to use to troubleshoot campaigns and make digital marketing better. I've been a digital marketer for 26 years; since '96, I've been in the trenches, looking at data and running campaigns, both my own and for clients. I was lucky to have free tools to use, like Google Analytics, because I was a small business owner and couldn't afford anything else. But because these tools were free, new features, bug fixes, and innovation were also slow at best.

Now that I've used Google Analytics for more than 17 years, I find that some things in GA actually made it harder for marketers to troubleshoot campaigns or get insights. For example, GA filters some bots -- like the IAB list of bots and spiders -- because they are required to. But the problem is that GA doesn't tell you which bots they filtered and how much was filtered. This leaves discrepancies that cannot readily be explained. In another simple example, by not having an hourly option in the most common report views, GA made it hard to see simple things like bot traffic occurring in the overnight hours or the same quantity of traffic every hour, which is a dead giveaway of bot activity. Of course there are workarounds that experienced practitioners can implement, but most people aren't advanced analytics practitioners or simply don't know how to set up those more detailed views; so they are left with analytics that amount to not much more than "scoreboards" which show nice quantity metrics like traffic, pages per session, time on site, etc. but not enough critical details to see fraud and bot activity.

I wrote up a how-to for setting up hourly views in GA and using those views to find obviously suspect and fraudulent sites referring traffic to your site.

This article has some more examples of what you can do with FouAnalytics to check, troubleshoot, and optimize your digital campaigns, that are not necessarily related to fraud.

Reach

Marketers have often told me that they went into programmatic media buying to "get more reach." They were convinced that there are millions of long tail sites that humans visit, and they could reach those audiences at scale on those tiny sites, instead of the large, mainstream publishers' sites. Well, let's test that assumption. What does the data tell us? Many of the largest advertisers spending millions in programmatic don't even get domain level placement reports, just end-of-month spreadsheets from their media agency. Those don't tell you that the "reach" assumption is wrong. I have done dozens of the following charts over the years to show advertisers that their campaigns end up on a small number of sites and apps -- i.e. they were not getting the large reach they thought they would get. In the 4 examples below, the top of each grid shows the total number of unique domains and apps their ads went to. But the details of each chart shows that the majority of the number of impressions were concentrated in a small number of domains. The rest of the "tail" had impressions as low as a single impression. That is not reach, especially consider the risk of bot activity on those long tail sites

Do you still you got lots of reach by buying through programmatic channels? Also, remember the example from 2018 where Chase reduced the number of sites showing their ads from 400,000 to just 5,000 and saw NO CHANGE in business activity? Right, 99% of those domains got so few ad impressions they didn't matter. We have not even mentioned the risk of ad fraud when allowing your ads to go to long tail crap sites. Here's an article to further bust the myth of "the long tail," a tall-tale sold by ad tech bro's to separate you from your money faster. See: https://www.dhirubhai.net/pulse/marketers-should-ween-themselves-from-long-tail-dr-augustine-fou-/

Frequency

FouAnalytics practitioners running programmatic campaigns should check if frequency caps were set properly or enforced properly by the ad platforms they are using. If you deployed in-ad tags to measure your ad impressions, in the FouAnalytics dashboard, you will see something called "fingerprint." A fingerprint is an anonymous representation of a unique device/browser combo, made from smashing together many javascript parameters and hashing it. The fingerprint table will show you the number of ad impressions that were shown to a device (a fingerprint) so you can easily see if your f-caps were set and working. In this example, they were not, because you can see nearly 200 ads served to the same device in the time period selected. Seeing these details allows you to troubleshoot your campaign and ask your agency to double check the settings and make corrections if necessary. You won't be able to see these details if you just get a spreadsheet with total quantities of ad impressions you purchased at the end of the month; so you might accidentally assume f-caps were in place, when they actually weren't.

Pacing

In FouAnalytics, the dashboard shows hourly data by default. Each green bar at the bottom third of the time series chart represents the hourly volume. FouAnalytics records visits in UTC (universal time code) and displays them in your local time zone. The examples to the right show the time series for a month, a week, and a day. You can zoom in by selecting the date range using the date picker at the top left; or you can click-drag-release to select the exact time period you want to zoom in on. The example here shows on-site measurement and it shows the natural ups and downs of hourly traffic. Note the overnight hours are lower in volume because humans are sleeping; and that makes sense.

By having hourly details you can easily see if something is strange or suspicious. In the following example, even without any color coding you can see something is strange since the hourly volume bars (bottom third) show the same amount of traffic every hour of every day. This is not a normal pattern of human visits to a site, so you know to investigate further.

The colored chart above shows the hourly measurement of ad impressions. What do the hourly green bars tell you? Right, that most of the volume of ads were show in the midnight hour (left 4 days) or between midnight and 4a (right 5 days). If all of your ad impressions for the day were used up in the overnight hours when humans were sleeping, you'd have little to none left to show during waking hours? Is this fraud? It may not be, but it is still wasted ad spend. By having this level of detail you can check if your media agency set up pacing correctly (spread impressions evenly throughout the day). In this case they didn't, so you can troubleshoot it with them. I might also suggest turning off the hours between 1am - 5am since few humans are awake and you are at the highest risk of bot activity if you run ads then. Google Analytics (daily view) won't show you these details; when you have these details with FouAnalytics you can check if there are problems and take action.

Where did your ads actually go (Breitbart)?

Some marketers might still be skeptical about using FouAnalytics. After all, they already get detailed placement reports from their ad platforms. But did you know that placement reports cannot tell apart spoofed domains from real ones? That's right, bad guys will lie about the domain in the bid request; instead of putting their own crappy/fake domain in the bid request, they put a more reputable domain so they can get higher bids. This is the domain that shows up in the placement reports. So the faked marthastewart.com inventory is comingled with the real marthastewart.com impressions; and you won't be able to spot these problems. By having the FouAnalytics in-ad tag in your ad impressions, we can check to see where your ad actually ended up. If it were marthastewart.com and it matched the domain passed in the bid request, then you're fine. If not, then something's wrong. In the screen shot below, we see a Mazda ad on Bongino and a Canon ad on Breitbart, despite those domains having been blocked. See this article for more salacious details about how bad guys do domain spoofing and get around your domain level blocks: https://www.dhirubhai.net/pulse/what-do-spoofing-dark-pool-ad-fraud-ad-fraud-researcher/

Who clicked your ads and arrived on your site?

If you just get click reports, how do you know which clicks were humans versus bots? I am sure that up till now, you have been very happy getting reports that showed lots of clicks and high click through rates from your digital campaigns. But should you count bot clicks as success? Of course not. I won't go into too much details here, since I wrote a whole article on bot clicks vs human clicks and how to use FouAnalytics to tell them apart. See: https://www.dhirubhai.net/pulse/full-funnel-outcomes-measurement-fouanalytics-ad-fraud-researcher/

Hopefully this was a useful overview of how marketers use FouAnalytics for more than just fraud detection. The additional details should help you check your campaigns for proper set up, troubleshoot issues that you can see, and take more precise actions to optimize your campaigns, that you couldn't do before with existing analytics and fraud detection. Common sense comes into play here too, once you have the right data and the right details to look at.

If you need any help or have any questions, as you use FouAnalytics, please let me know. I am here to help and I am a fellow practitioner. I also tune the bot and fraud detection algorithms of the FouAnalytics regularly, and look at data every day, as I have for the last 10 years. Happy (fraud) hunting, y'all! More importantly, happy optimizing your digital campaigns with new details you didn't have before! ;-)

Further reading: https://www.dhirubhai.net/today/author/augustinefou