Why Programmatic Advertising Is Safer Than You Think
When it comes to programmatic advertising, it’s easy to become preoccupied with its perceived constraints. After all, it’s proven to be highly disruptive — and, of course, negative news always has a way of making its way into headlines. But is programmatic inherently problematic? It certainly has presented marketers with some new challenges, but in many regards programmatic can be just as safe — and much more effective — than legacy options.
Thanks to advances in the realm of artificial intelligence (AI) and machine learning (ML), programmatic is becoming even more safe and effective by the day.
Understanding Current Preconceptions and Realities Around Programmatic
While some of the more negative headlines out there around programmatic are a tad hyperbolic, of course programmatic is not totally faultless. For starters, issues around brand safety didn’t really exist in the pre-programmatic world.
It’s not entirely coincidental that the global rise of programmatic advertising has corresponded with ever-rising rates of ad fraud such as Invalid Traffic (IVT). Fraud, brand safety and a lack of transparency were cited as one of the top problems with programmatic in 2019 research reports from IAS and Digiday.
Highlighting a More Complete Programmatic Picture
These two reports showcase a very real reality for marketers today: programmatic can add complexity to media buying, and it is leading to potentially unsafe conditions. But is this the whole picture? Not quite.
If programmatic was so problematic, then why does eMarketer predict that programmatic ad spending is expected to reach almost $56 billion in the U.S. by 2020? The answer, of course, is that it works.
The same 2019 research from Digiday showed the other side of the coin too: over half of those surveyed said programmatic “increased targeting and optimization” while around one in four said it provided greater efficiency. This matches the findings of a report from Forrester and PubMatic, which found that more than half of global media buyers polled seeing “better audience targeting” and “more effective customer engagement” from programmatic; an additional 43 percent said programmatic leads to “better reach and frequency caps.”
However, not all programmatic spending is equal. In particular, both deal type (open exchange buying vs. programmatic direct vs. private marketplace deals) and final hosting environment (browser vs. app) can dramatically change the equation.
Let’s dive deeper into the differences in devices. According to 2018 figures from IAS, observed fraud rates were significantly higher on desktop for both programmatic display and programmatic video than they were on mobile. This aligns with research from Smaato and Protected Media, which found fraud to be far more prevalent in mobile web (i.e. browser) advertising than in-app advertising.
Like anything else in life, it’s all in the company you keep. For example, channels certified by the Trustworthy Accountability Group (TAG) are 84 percent less prone to fraud, while major U.K. newspaper The Guardian was able to reduce its risk of video advertising-related fraud by 72 percent thanks to the IAB’s ads.txt initiative.
How ML Improves Programmatic Even Further
Of course, just because programmatic is safer than many think doesn’t mean there isn’t room for improvement. As long as fraud and brand safety are around, everyone in the ecosystem needs to work together to improve programmatic buying and selling.
But, there’s a lot to be really excited about in the realm of AI and ML. How can AI improve programmatic buying?
One area to be especially bullish on is the application of Generative Adversarial Networks (GANs) to improve our ability to spot and fight fraud. With GANs, two AI models are built: one designed to perpetrate fraud and one designed to fight fraud. They’re then pitted against one another and through ML aim to try to continually one-up the other. These deep learning algorithms adaptively learn complex patterns and extrapolate to detect newer types of fraud in real-time. This way, it’s possible to more effectively predict and prevent many possible instances of fraud, and have a way to determine how fraudsters will likely be evolving in the future.
In addition to GANs, the industry is moving forward on the following:
- Supporting partners working towards real-time content analysis for improved brand safety.
- Discarding bot-clicks through pattern matching.
- Using dynamic threshold based techniques to detect known types of ad fraud post-click, among other actions.
- Developing standards on behavior, so anomalies can be more quickly spotted and detected.
- Leveraging mobile-first signal triangulation to verify the validity ad clicks/ad actions.
- Improving dynamic creative optimization (DCO) to make AI-powered personalization a reality.
- Identifying bots and scripts through pattern analysis of impressions and clicks in real time.
- Improving computer vision algorithms designed to help spot faulty/spoof creatives before they run.
Calculating Programmatic’s Future
When it comes to programmatic, it can be tempting to approach it with a Manichean worldview: Either it’s all good or all bad. The reality is somewhere in between. Is programmatic safer than you might think? Yes. Is it highly effective? Definitely.
Is it totally fault free? No. But, thanks to the efforts undertaken by industry players such as DoubleVerify in applying AI and ML to further improve the programmatic ecosystem, it’s becoming safer and more effective by the day.
Don’t let the fear-mongering headlines get to you. Programmatic is far less problematic than you think it is.
Related story: Google Programmatic Changes in a Couple Months