Advertising fraud has been a consistent problem for the digital marketing industry, consuming approximately $19bn of ad spend in 2018 alone, but the issue continues to grow year-on-year. In the last five years, ad fraud has changed from an inevitable consequence to a main topic of conversation in the industry.
The traditional approach to ad fraud has been to publicly expose new fraud methods by implementing industry standard measures or benchmarks. Typically, by the time these methods are caught and publicised, a new method of fraud has already begun. While the discovery and removal of ad fraud has improved for clients, these reactive measures aren’t dealing with the cause. The industry needs to transform its tactics and position preventative actions at the centre of their strategies.
Everyone knows your benchmarks
In the past, marketers have been overconfident about the lasting impact of their fraud rules. A prime example is the industry's reaction to two of the first fraud types to be publicised and tackled by app advertisers - click spamming and click injection fraud.
In 2016, click injection evolved from click spamming and became a more sophisticated way of capturing installs. As click injection became visible, many app advertisers reacted by implementing generic data techniques to detect click injection such as a click to install time benchmark of 10, 20, or 30 seconds.
Implementing this methodology resulted in the rejection of a sizable number of fraudulent installs leading to the eventual reduction of rejected installs. This drop can be viewed as a measurement of success but, in reality, fraud methods had adapted to evade these preventative measures. Marketers can never become complacent; they need to stay vigilant to other ad fraud opportunities.
Be prepared to expand your rules
While your numbers may look good, it might be because you haven’t identified all the types of fraud and need to revaluate your activity. Last year this was evident in Google’s annual review when they identified the existence of click fraud – changing the real value of ad spend.
In Google’s Android Security & Privacy 2018 Year in Review they celebrated the steps they have, and will be taking, to protect Android devices and users. A key call out was the yearly increase in potentially harmful apps within the Google Play Store (+100% YoY).
This is directly linked to the introduction of a Click Fraud categorisation for harmful apps by Google. Prior to this, click fraud was not considered by Google, and their install rate for harmful apps had halved. In fact, if we remove click fraud from these stats, the data shows that the install rate of harmful apps on Google Play declined by 31% year-over-year.
The key point here is the evolution of Google’s benchmarks and measurements at the detriment of showing the world improved figures. Decreasing your fraud numbers isn’t always a measure of success, it may be a sign that it’s time to evolve existing rules and create new ones. We need to ensure we follow by Google’s example in order to proactively tackle the evolution of fraud.
New Data Points are being ignored
Click and install time measures have been the predominant focus for tackling click injection fraud, so much so that other opportunities to combat fraud are being ignored by marketers. The recent release of the Google Play Referrer API for attribution platforms provided a new data point for advertisers to tackle Click Injection through install verification. Despite this data point being both informative and definitive, the industry has failed to effectively adopt it.
Retaining a stagnant approach to fraud like this makes the playing field much easier for fraudsters to earn revenue from your investment. Understanding how to evolve your existing rules is key. New data points and technology is always being developed, making it easier to validate and evolve your existing rules.
Think outside of the box
With a wealth of data at your fingertips, it’s easy to become overwhelmed and focus on the data you are familiar with, while deeming all other data points to be white noise. A lot of the rules and metrics we scrutinise are focused around time, frequency, and location of the occurrence.
If we have a universal understanding that these data points are also recognised by fraudsters as the go-to metrics for policing fraud, it’s easy to realise that these data points are more likely to be manipulated. With more datapoints available than ever before, it’s important that we think outside of the box when it comes to data. Ad fraud is hiding in the data points you don’t expect – you have to think like a fraudster to find it.
Here are a few key tips on proactively tackling fraud:
- Organic data is key – Use your Organic data as a comparative variable when testing your new rules or trying to understand your consumer behaviour. This will give you a truer understanding of the trends you should be seeing across a clean paid media data set.
- Your rules are hypotheses – Your rules are not set in stone and you should stress test them as much as possible before applying them to your paid media. Predict the impact you should see from your rules and understand the results you are seeing. If they vary significantly, take some time to understand why.
- Track individually – Measuring the overall fraud on your paid media investment may make you lose sight of the bigger picture. The introduction of new rules will always bring this up. Each rule should be tracked individually to understand where the problem areas are across your activity.
- Common sense – Your general knowledge is your best friend. Keep an eye on your trends. If there are newer app versions or the latest iPhone or Android phones being released, ensure that you’re seeing these trends within your data. If you aren’t, begin to question them.
To learn more about how we combat ad fraud, get in touch.
Micarla Joseph, Senior manager – Buying at M&C Saatchi Performance.