PERFORMANCE MARKETERS use FouAnalytics to reduce CPO/CPA

I have written over the years that performance marketers should be more careful with their ad spending and not assume that they are "immune" to ad fraud.

For reference from 2019: Performance Marketers, You're Still Getting Ripped Off by Fraud

Performance marketers have been led to believe that because they are optimizing for "performance" already, they don't need to worry about ad fraud, because they are seeing "performance." Let me unpack this circular logic a bit. Some marketers use clicks on their ads to mean performance. Some marketers use sales or orders to mean performance. Let me show you how both can, and are, still being tricked by fraudsters to make money fraudulently.

Clicks as "performance"

This first one is easy to prove wrong. And more marketers are realizing that using clicks for performance is not a good idea, because it is so easy for bots to simulate clicks. You may have realized this because you looked at your own Google Analytics and saw nearly 100% bounce rate or nearly 0 seconds time on site from your paid digital campaigns.

These visitors that clicked through from your ads are not valuable, because they didn't do anything on your site, and they didn't stick around. They left right after they arrived. That's what bots do. And that's what you see in the 15 donut charts below. Orange and red mean bots, and dark blue means humans. These are clicks from utm_source=programmatic. Nearly all of the clicks were bots clicking on the ads. When they get to your site, they leave right away so they can move on to create the next fake click.

Google Analytics doesn't show you this. And the legacy fraud verification vendors still tell you that ad fraud is 1% (and has been for the last 8 years). Clearly they are missing most of the fraud and bots. When you have FouAnalytics on the landing page, you can see the quality (or lack thereof) of the clicks that arrive on your site. You can also track back to where the bad clicks came from and make adjustments to your campaign set up, including turning off spend.

CPO "cost per order" or CPA "cost per acquisition" as performance

A lot of marketers will now say "we have done away with measuring performance using clicks." Oh really? "Yeah, we optimize for CPO ("cost per order") or CPA ("cost per acquisition"). So we're immune to fraud because bots don't buy stuff."

Let me stop you right there and refute that. Bots can, and do, convert. See the data and evidence in this article from March 2024: Performance marketers, take note -- bots convert Bots also complete lead forms with valid (stolen) data so they can earn the CPL ("cost per lead"). What's easier to understand is that bots bought up most of Taylor Swift's tickets within 15 minutes of them going on sale, so that scalpers could re-sell them for 5X the cost on the aftermarket. And bots have been buying virtual goods using bitcoin and stolen credit card numbers to launder criminal proceeds for at least the last decade.

Now that you understand bots can buy stuff and trick other forms of "performance" marketing like CPL or CPI ("cost per (app) install"), I can tell you that that's not how they impact your paid digital advertising campaigns. They are not buying something from you. They are tricking the attribution so optimization algorithms allocate more ads and budget to the fake sites and away from real sites with real human audiences. You ARE getting orders and conversions, but those orders and conversions were not caused by the ads shown to bots. It's the attribution models being tricked into attributing the orders and sales to these channels. So the CPO ("cost per order") you see now is far higher than it should be, because you are paying for ads that are shown to bots when you didn't need to. Those orders would have happened anyway.

Two scenarios that are still tricking performance marketers

If you're still with me to this point, let me re-articulate with more detail the 2 scenarios in which fraudsters can still trick performance marketers and get away with ad fraud. I'll start with the easier one.

1. cookie stuffing to mark browsers/devices/users as "exposed" -- human users convert -- i.e. buy stuff. Advertisers want to know which users bought from them after they saw their ad -- i.e. were exposed to the ads. So what do you think fraudsters do? They mark as many browsers/devices/users as possible as "exposed" so when any of them buy stuff from you, they can claim credit for having driven that order/purchase. This is exactly the same scheme as affiliate fraud, which has been rampant, as documented over the last 2 decades by the great Ben Edelman In essence, as many affiliate cookies as possible are stuffed in as many humans' browsers as possible, so when any of these humans actually convert, the affiliate revenue shares are paid out, unnecessarily, because the humans would have converted anyway, without seeing the ad or visiting the affiliate's sites. This means the advertiser over-spent on affiliate marketing and the cost per order was higher than it should be, if this kind of affiliate fraud were eliminated.
2. tricking the optimization algorithms in performance campaigns - advertisers running performance campaigns on platforms like PMax (performance max) are also easily tricked into paying for ads they don't need to -- i.e. that didn't cause orders. The PMax optimization algorithm uses signals that it can easily detect -- clicks on the ads and action on the landing page (as measured by a conversion pixel). In the FouAnalytics chart below, you can see the clicks from utm_medium=pmax. These clicks were mostly from bots (84%) dark red. On top of that, these bots simulated clicks on the landing page (click charts on the right). The combination of clicks on the ads and clicks on the landing page (simulated engagement) are enough to trick PMax algorithms into allocating more ads and budget to the fake sites and apps that use these kinds of bots. The advertiser sees orders during the same time period when these performance campaigns were in-market. So they mistakenly attribute those sales to these performance campaigns, even though these bots were not actually the users that converted. So the advertiser is over-paying for ads that didn't cause the orders/sales/purchases they observed. That's why I said earlier that the CPO ("cost per order") is higher than it should be, if this form of performance fraud were eliminated.

So what?

Hopefully you now understand why I said earlier that performance marketers should still be more careful and not assume they are immune to bots and fraud, simply because they are optimizing for number of orders or cost per order. The best way to see if you are affected by this kind of fraud is to use FouAnalytics to measure the ads (where possible) and measure the landing pages, so you can "see Fou yourself" whether bots are clicking on your performance ads and clicking on your landing page to trick the performance optimization algorithms into allocating more of your ads and budget to fake sites and apps.

Happy Sunday Y'all!

More screen shots and examples from FouAnalytics: https://www.dhirubhai.net/today/author/augustinefou