Performance marketing actively uses data feedback to optimize a campaign, both in terms of impact and expense. Measuring and analyzing actions like clicks, leads and sales makes it possible to refine a campaign, focusing on what works and improving (or eliminating) non-performing features. Which search keywords worked? What geographic areas produced results? At what time of day? Which genders responded? How old were the people you reached? Which placements earned clicks? Tweaking parameters in a campaign allows you optimize the performance and improve cost effectiveness. The result will be a better focused, better performing campaign.

For best results, the process is repeated often, with data feedback informing refinement to any and all of the parameters that impact performance in the campaign.

This blog post aims to illustrate what’s really happening in and around a campaign, both in terms of drivers that marketing controls and the outside factors that interact with the campaign to affect its real success. We’ll explain how to identify problems in a campaign, and how – if armed with data and an understanding of the ecosystem – these problems can be solved to optimize marketing expense.

Sounds like a battlefield, right? Well, in the spirit of the imminent Super Bowl, let’s liken it instead to a game of American football!

To visualize a campaign and its clicks and leads, we’re going to model them across the 100 yards and two endzones of a football field. In the diagrams that follow, clicks are represented as circles and leads as stars. You want to reach real sales from real people, right? Yes, so the objective is to get into what we’re calling the Human Endzone. And the closer you get to that human touchdown area, the more likely you are to achieve real conversions to sales. A well-constructed campaign is optimized to get there.

In the analysis that follows, each illustration will contain the exact same playing field so it’ll be easy to compare the outcomes. We’ll lay out competing scenarios, first, and then compare the corresponding numbers. As you’ll see, proper fraud detection is a vital part of the strategy to reach the endzone. Ready? Set. Hike!

Campaign without Fraud Detection
The campaign without any fraud detection is shown to the left. Starting at the bottom in what we might think of as the game’s kickoff we see a campaign starting without optimization. Here you have already set your target audience according, let’s say, to gender, age and geo-location and maybe, if applicable, search keywords. The illustration shows that clicks and leads are generated throughout the field, and they are evenly distributed. The only metric you have is the likelihood of conversion.

In the next image, which corresponds to the First Quarter, you start optimizing to eliminate the worst performing parts of the campaign. That will eliminate some clicks and leads but as they’re the ones furthest from the endzone they probably wouldn’t have been converted anyway.

Through Q2, no leads and clicks are happening in the eliminated parts of the field, and again that’s a good thing! At the same time, the middle of the field has become more crowded, with more clicks and more leads! Unfortunately, from here you still need a Hail Mary to get in the endzone! Or in our jargon, this part of the field underperforms compared to the area closer to a touchdown, so a new optimization is required.

Another round of optimization in Q3 eliminates the middle part of the field and thus the least performing clicks and leads of the campaign.

By the Fourth Quarter, all the clicks and leads have moved to the tiny area still in play. By the final whistle any further optimization would only lower the volume and not increase the quality.

Unfortunately, once the campaign has finished and its results reviewed, it will be clear that the campaign performed poorly, with each optimization having increased the average Cost per Acquisition (CPA) and Cost per Click (CPC), where a better performing campaign would have seen them fall.

Campaign with Rudimentary Fraud Blocking
The same campaign is considered again, but now we have added the ability to see which clicks and leads are generated by bots, click farms and sweatshops. The illustration at the bottom shows again that clicks and leads are generated throughout the field at kickoff, but many of them are now tagged as fraudulent (as marked in red).

As this game progresses, you realize you have a more cunning opponent – or a team of them! Now as steps are taken to optimize the campaign, the other side changes their tactics, too, which is exactly what happens when bots, click farms and sweatshops adjust their strategy in order to continue to generate clicks and leads that earn them their bucks.

In Q1 you successfully eliminated most fraud, not by detecting it and filtering it, as it were, but by narrowing the campaign to eliminate it. While this means you have less fraud to pay for, it also means you have less opportunity to reach your intended targets.

By Q2, another problem emerges, namely that the fraudsters have adapted their game plan, and have returned more advanced than before! Fraudulent clicks and leads have reappeared and moved to the middle part the field. This has affected the statistics of this part of the campaign and after further analysis you again narrow the campaign to eliminate the noise.

Another optimization in Q3 eliminates most of the fraud but also dramatically reduces the scope of the campaign. And by the final whistle, the fraudsters have infiltrated the campaign again, and are occupying the area closest to human endzone. But will they ever convert to a sale? No, of course not, but they will get affiliate attribution fees.

This is the endgame for most on-line campaigns at present, with bots and other nefarious forces mingling with human traffic, and doing so with such sophistication that they can only be spotted in the earliest stages. All of which makes for under-performing campaigns, plagued with fake leads and even the risk of litigation.

Campaign with Intelligent Fraud Detection
The same campaign is again shown at the bottom, but now the marketing team have proper fraud detection in place from the start of the campaign. Armed and are able to see which clicks and leads are generated by bots, click farms and sweatshops.

As before, at kickoff we see clicks and leads generated throughout the field, with fraudulent clicks and leads identified and marked red. As you’ll see, in this matchup, the game plays out very differently.

Why? Because intelligent fraud detection allows you to focus on your human interactions, and to optimize with them in mind, without narrowing the campaign to avoid the other noise. What about the bots and click farms, you ask? Well, you don’t have to worry about them because most affiliates accept when you refuse to pay for fraud.

But you have to have implemented proper fraud detection to be able to reliably identify fraud.

Play commences and throughout the first half you optimize to remove underperforming human traffic. In Quarters One and Two, underperforming humans get removed from the campaign. Even as bots reoptimize themselves to the evolving campaign, they remain irrelevant and cost-free.

Reoptimizing again in the Third and Fourth Quarters, the impact is even clearer. Undistracted by bot traffic, your campaign reaches the widest possible opportunity set, with the human endzone and conversion to sales immediately ahead. The last stage shows a very focused campaign with a high percentage leads compared to the number of clicks, and a high probability of conversion to a sale.

The illustration also reveals that the bots, click farms and sweatshops have not evolved, because they only evolve out of necessity.

Comparison of the Two Campaigns

The table below compares the first described campaign without any proper fraud detection and the third campaign with proper fraud detection. The table will show only the number of clicks and leads you are paying for. The table only contains overall numbers without the metric of likeliness. Against the backdrop of the illustrations, the numbers below show clearly that the ability to optimize your campaign on your human audience only – without also optimizing to eliminate bots – does greatly improve performance.

The table shows both situations side by side and each cell contains the number of clicks and leads displayed on the field. The left part of the table shows the campaign in which you are not able to detect fraud and resembles the grey clicks and leads. The right part of the table shows the campaign with proper fraud detection enabled and by eliminating bots from the optimization process the campaign is optimized on its human audience only.

You can clearly see that even without the likeliness of conversion metric, the right campaign is performing much better. The left part starts at Kickoff with 12 leads/ 50 clicks, which is a ratio of 24%. After multiple stages of optimization, the left part ends at the Final with 8 leads/ 30 clicks, which is a ratio of 26.7%. The right part starts at Kickoff with 11 leads/ 40 clicks, which is a ratio of 27.5%. It ends at the Final with 7 leads/ 18 clicks, which is a ratio of 38.9%. This clearly shows that removing bots from the equation makes a campaign much more performant. It also shows that on the left part of the table round 05 contains the maximum possible performance. Any further optimization will remove only interested humans and thus a huge jump in the number of bots relative to the number of humans causing the performance to go down the drain.

Without proper fraud detection you will also not be able to scale without affecting your performance. You will only get traffic that wants you (your affiliate commission) instead of traffic you want. This means when you want to grow you will have to implement a proper fraud detection. You might be thinking about additional costs for proper fraud detection? If you refuse to pay for fraudulent traffic you will save money, and the best part is: We know that our clients save more than our fraud detection costs.

What is the takeaway? That campaign performance can and should be optimized to improve performance, and that can be accomplished without incurring fraud-related costs. How?

– Robust fraud detection means you can refuse to pay for fraudulent traffic
– The savings of fraudulent traffic outweigh the costs of fraud detection
– By optimizing a campaign on human traffic only, it efficiently reaches the widest possible target audience.

– Have a real proper fraud detection mechanism at your landing page which only detects and does not block, preventing fraudsters to A/B test your fraud detection.
– Have a fraud detection that reports the fraud status per lead in real-time to your back end
– Having a real-time fraud status allows you to follow-up on human generated leads only.
– Removing fraudulent leads lowers litigation risks as fraudulent leads are not TCPA compliant.
– Sample within the group of fraudulent leads to verify that fraudulent flagged leads are really fraud
– Create proof of consent for each submitted form as evidence in case of a dispute.

If the performance of your campaigns does degrade over time and you are wondering why that happens, you might ask yourself: Do I optimize my campaigns incorrectly? Or is my current fraud detection tool missing the continuous evolving fraudulent traffic?