Quantifying Click Fraud vs The Black Magic of Machine Learning?

Cheers to another insightful session brought to you by ClickGUARD. Our brilliant host, digital marketer Jason Pittock, and the King of Data Miloš Đekić are tackling a topic you’ve been buzzing about: how can we block bad traffic and wasteful clicks with a transparent solution?  

This topic has been on our priority list for a while and this Reddit thread confirmed the need for addressing the issue NOW.  

Let us be frank right off the bat: when it comes to stopping click-fraud, blindly trusting in machine learning can take you no further than believing in black magic. You may give it a shot and wonder at the results. 

We think you know better than that, rather looking for some tried and true solutions. That’s what this webinar was all about, showing once again what ClickGUARD can do for you. We’re all about transparency and it shouldn’t be otherwise when it comes to mitigating “all those money-wasting clicks that break people’s banks,” as Jason accurately defined click-fraud. 

“The first step to blocking bad traffic and wasteful clicks is QUANTIFYING”

If you’re anything like Miloš, you might have had a Reddit account for a few years but haven’t really dug in. Well, you’re missing out quite a bit. Unlike other platforms, there’s not much, if any toxicity going on Reddit (Miloš is so hyped about it that he “had to sing a little praise for Reddit”). 

And if you’re reading this, you will probably be stoked about some PPC and digital marketing subreddits, full of “gold nuggets''. One of them is this thread started by a cybersecurity engineer, who was wondering if click fraud is a real problem and if people actually want to pay for solutions to put an end to it.  

As it turns out, our fellow marketers and cybersecurity engineers are well aware of the problem but seem timid to engage solutions. 

Do you feel the same way? Is it because of the lack of transparency and those outrageous claims about “fraud” and “competitors clicking on your ads”? 

We can’t blame you, and we can’t agree more with the user who is seeking “a solution that can somehow bring transparency to click fraud detection… that would effectively spell the end of it.” 

If you didn’t know, Miloš lets you know: this is what ClickGUARD does. 

Side noteIf you feel a bit sheepish in commenting and participating in Reddit threads - as Miloš is despite having plenty of knowledge to share, consider taking his approach and earn some good Reddit karma. 

Be your own hero, don’t “sit back and relax”

We’re not saying you’re being manipulated, but if you believe that AI combats click fraud while you “sit back and relax,” you might as well call that robot your Prince Charming. 

Your Charming Robot may not have a white horse, but you can recognize him by the wild “fraud rates,” and a special way of “transparently” acknowledging false positives (with false-positive rates). 

Before you know it, you’ll be presented with a “proprietary tech,” which, would you believe, keeps your website safe and fraud-free. Too good to be true? Yes, indeed. 

“You’re busy setting up your campaigns, tweaking stuff. That’s no excuse to put blind faith into a black box doing “something” with your campaigns.” 

Unfortunately, people do fall for it, and unsurprisingly, some of the biggest names in the industry get called out for being fraudulent themselves. 

Want to give it a try? Arm yourself with patience, ask some questions, and see what happens. We’re not fortune tellers, but we can predict this much: 

You: WHY did you block this IP range? 

Them: We’ve determined that clicks are from your competitor.

You: Ok, HOW did you determine these clicks are from my competitor? 

Them: Well, you see, we have this intelligent proprietary algorithm that… something, something. 

Their bottom line: no need to think or worry, the robot does it for you. 

Translation: You have to blindly trust what the fancy-looking reports say. ZERO TRANSPARENCY. 

How do marketers accurately quantify click fraud, though?

This is the real question to which the King of Data has not only answers but also real solutions. 

Say you or someone you know decided to trust an ad vendor who is actually respecting the targeting settings and reaching the right audience. 

However, something is clearly off, as the well-researched audience doesn’t convert. It might mean very little that one of the greatest retailers of all time has been through this, back in the late 1800s. 

“I know that half of my advertisements are waste, I just don’t know which half”. 

John Wanamaker 

Admittedly, a portion of the reached audience is bound to never convert. But, if Wanamaker didn’t have the means to find out which half is waste, you do. 

A simple yet efficient way to identify waste. 

Employ click forensics - the science of analyzing clicks and post-click behaviour. It brings insight into the quality of the traffic you’re paying for (your ad vendor will not bother to analyze or show). 

You will learn things like:

• What kind of sources has been sending you clicks: 

1. Were there any IPs or IP ranges previously involved in nefarious activities or with a history of clicking but never converting? 
2. What about ad placements, are you getting a ton of traffic from mobile games that never convert?

• What kind of post-click behaviour indicates that there will be a conversion? 
• What kind of interaction with your site means you should not expect conversion?

Click forensics gives you a way to identify what is a waste for you, for your business, for your client. There is no universal data for this. 

Hint: services saying “we know your industry so apply these rules to get results'' are off track. 

Not a single ad campaign is the same as another, keywords are different, sometimes the quality of your traffic fluctuates between different countries, ad groups… anything really. 

Exercise: How to actually quantify waste (and by deduction, click fraud). 

Focus on a single campaign and look at the forensic reports. 

Check: 

• How many ad clicks you’ve gotten within a single month from nefarious click sources that didn’t convert;
• How many clicks you’ve received from mobile apps that didn’t move the needle for you (let’s say they spend little time-on-site and don’t interact). 

Multiply these numbers by your CPC and you’ve just quantified waste. 

Follow-up steps: 

Make this data sing (read “informative”), then make it work (read “actionable by setting rules to rid yourself of waste”).

AI and machine learning work. Yes, but there is a catch.

For good reasons, Jason is a sceptic when it comes to ad vendors’ promises to provide you with results generated through “AI magic.” We’re all for smart solutions (like cybersecurity with a minimum of effort), but buying into a one-size-fits-all approach doesn’t fall under this category. 

We’ve covered this topic in our session about False positives, showing you how you could miss out on actual customers (throwback to webinar #5). It’s worth mentioning once again that ClickGUARD can easily keep you safe and be in control, with UNPARALLELED transparency. Unless you’d rather spend a fortune on smart campaigns with no paper trail... 

“The goal for many of these black-box services is to get the buyer to believe they are “protected”. You’re buying a piece of mind.”

Many of these services do work and generate some results (we’ll learn about the why’s and how’s in a future session, Miloš assures us). But here’s the catch: 

- You don’t really KNOW that the potentially better results are generated by their actions.

- You often don’t know what exactly goes on and there is no way you can get a paper trail. 

- You could be hindering your results because the supposed “AI” is applying some simple industry templates to your ads in the back, and you can’t even know it. 

These are very enough reasons to fear that you will prevent good, legitimate customers from seeing your ads by employing click fraud mitigation tools. To set your mind at peace, ask your vendor as many questions as possible, until you understand what’s going on. 

Does the vendor claim that  “30% of your ad traffic is coming from your competitors”? 

You know that this is unlikely, so ask them for the paper trail. 

No paper trail or data? 

Ask them to demonstrate HOW they’ve determined that your competitor has clicked on the ad. 

Click Forensics for the win. 

“You can’t do anything if you don’t know what’s going on, and you can’t know what’s going on without click forensics. 

The insight you get into the quality of your paid traffic is irreplaceable and you get to pinpoint and quantify waste, which not only answers the question “which half is wasteful” it also gives you the opportunity to do something about it.”

For the full conversation, inside jokes, the bonus question, and the witty answer, tune in to the audio version of our 7th Webinar session, “Quantifying Click Fraud vs. The Black Magic of Machine Learning.”