Digital Campaigns Measured by FouAnalytics

Over the last 5 years, FouAnalytics practitioners have run Facebook campaigns side by side with programmatic display campaigns. In the examples below I will show you how to use FouAnalytics to measure the quality of clicks coming from Facebook. We were fortunate to also have reliable conversion events to analyze. Since Facebook does not allow us to measure the ads themselves, we put the FouAnalytics tags on the landing pages to which the clicks go, so we can see whether the clicks are from humans or bots.

When you run Facebook campaigns, be sure to turn OFF Facebook Audience Network, because that is where most of the ad fraud is. When you set up a campaign, look for the "placement" section and select where you want your ads to go. On Facebook, Instagram, and WhatsApp are fine; be sure to uncheck the "audience network" checkboxes.

The slide below from 2017 shows four paid display sources. "Display 1" was Facebook paid display ads. Note the total number of clicks, 8,482; 2,036 were humans (24% dark blue). Of those humans, 818 completed the conversion event (note the dark blue in the conversions donut chart). This shows that not only do Facebook campaigns drive more humans, it appears to deliver more relevant users. Note the much higher human conversion rate -- 40% compared to 5 - 9% from the other sources. This tells me the humans that clicked, wanted to be there, and were more likely to convert.

The next slide (below) is another example from 2018. If you look down each column, you will see arrived, clicked, converted. Note that Facebook, Sizmek, and Basis all drove similar quantities of clicks to the site ("arrived"). But note the drastic differences in quality -- the percent of dark blue (humans) vs dark and light red (bots). Note that "clicked" means the users that arrived on the site took some action. Note that bots might arrive on the site but they don't take further action; so most of the donut charts in the clicked row are light and dark blue. Finally, humans convert, bots don't. So in the bottom row - "converted" - you can see the number of conversions, and most of these were dark blue (humans).

I never ask my clients about the CPMs they paid. But based on the details in the slide above, they can easily compute the cost per HUMAN conversion, since they know the total dollars they spent, and the corresponding CPM. By eye-balling the slide above, you can surmise that Facebook display worked well for them, and yielded the lowest cost per conversion, compared to the other programmatic display campaigns.

If we generalize further, you can use FouAnalytics to measure clicks arriving from all paid sources. The chart below shows 3 paid display sources, 2 paid search sources, 2 paid social sources, and 1 native. "F" is Facebook. Lots of dark blue clicks arriving on the landing page. You can do your own cost per HUMAN-click calculations and decide how to allocate, or re-allocate your budgets across all these sources to optimize your cost per human click (hCPC). If you have the ability to re-allocate budgets across channels, you can allocate more budget to channels sending more human clicks and less budget to those channels sending more bot clicks (dark red).

Practitioners will understand that even if you got a lot of clicks, many of those clicks are not human clicks. See the TikTok example above. Bots are programmed to click on ads to make them appear to be performing. If you don't have a way to tell apart which clicks are bots versus which clicks are humans, you can't take action. That's been the problem with existing invalid traffic detection. Not only are they very bad at catching the bots, they also do not label humans (blue). So there's not much you can do with a percentage IVT reported to you in a spreadsheet or dashboard. Some may be surprised to see Instagram with no bot clicks. Note that Instagram has many fake accounts; but because fraud bots cannot get paid for ads that run on Instagram, they are not actively trying to cause ads to load or click on them while on Instagram. Automated accounts on Instagram are used to like and share posts -- for example, to amplify and spread disinformation. In this campaign example, the clicks we got from Instagram were human clicks.

To provide a contrasting example of clicks coming from programmatic, we can isolate "utm_source=programmatic" in FouAnalytics from 6 different campaigns. We can see the quality of clicks coming from programmatic channels, we see a lot of orange (declared bots) and red (bad bots) and very little dark blue (humans).

If you take the dark blue percentages -- e.g. 1 - 5% -- you will realize that to get those humans to click, you'd essentially have to buy 20 - 100 times more ad impressions. That costs you 20 - 100X more than if you allocated your budget to paid media channels that demonstrably send more human clicks. Put another way, you can buy far fewer ads, pay higher CPM prices and get better outcomes. That's because you're showing ads to humans. If you pay on a CPC basis, you definitely need better analytics to know if the clicks you are getting are bots or humans. If you pay on a CPC basis, the bots are motivated to click, otherwise they won't make money.

One final thought, some marketers will be asking "what if we are not performance marketers, and we don't really track click throughs to landing pages?" Well, obviously you still need to show your ads to humans, right? Many humans remain logged in to Facebook, Instagram, WhatsApp all day long. Show them your ads (be sure to turn OFF FAN to minimize the obvious fraud). Many humans are logged into Chrome, Android, and Gmail all day long. Show them your ads (be sure to turn OFF Google search partners and Google Display Network to minimize the obvious fraud). There's a way to do good digital marketing. Look closely at analytics and outcomes. Be sure you can tell apart bot clicks from human clicks. FouAnalytics is ALWAYS free for advertisers to use on websites. If you want to try it, message me.

Read another Facebook example that I wrote up: Slow Digital Marketing, Like Slow Juicing, Works Better