How Does VeraViews Use Artificial Intelligence & Machine Learning?
Ad fraud, which includes bot traffic, click fraud, and many other fraudulent activities, is a pervasive problem that has been plaguing the advertising industry for years. Online advertising fraud resulted in the loss of more than $80 billion in ad spend in 2022 alone, and it’s projected to grow to over $100 billion in 2023.
However, recent advancements in artificial intelligence (AI) and machine learning (ML) have spurred the innovation of essential tools for identifying and preventing advertising fraud in all its forms. These technologies enable advertisers to detect and prevent fraudulent activities in real-time, ultimately saving them significant amounts of money and safeguarding their reputations in the process.
By leveraging artificial intelligence and machine learning businesses can take control of their ad spend, ensuring that it’s spent effectively and their ads are reaching real human audiences not just bots.
In this article we will discover the advantages artificial intelligence and machine learning offer for ad fraud detection, explore their use in the VeraViews adstack and take a closer look at how VeraViews uses a three-phase approach to eliminate fraud.
What are the advantages of AI and ML for ad fraud detection?
One of the most significant advantages of AI and ML based systems is the ability to detect anomalies. These technologies can analyse large amounts of data and identify patterns and behaviours that are abnormal or out of the ordinary in a manner that is vastly more accurate and time-efficient than manual detection or other previous methods.
This can be used to identify fraudulent activities such as bot traffic, click fraud and other forms of ad fraud. By detecting these patterns, advertisers and publishers can take action to prevent these fraudulent activities from occurring or stop them going forward if they have already occurred and protect their advertising budgets.
Predictive analytics is another powerful tool that AI and ML offer. By analysing data, AI and ML can predict which ad placements are likely to be fraudulent. This can help prevent ads from being placed on fraudulent websites or in fraudulent apps. By analysing data from previous ad placements, these technologies can identify patterns that are likely to indicate fraudulent activity. This can help advertisers avoid these placements and protect their advertising budgets.
Fraud modelling is also an effective use of AI and ML for preventing ad fraud. By building and training models that simulate fraudulent activities, advertisers can detect and prevent fraudulent activities in real-time. These models can analyse large amounts of data and identify patterns that are indicative of fraudulent activities, allowing advertisers to take immediate action to prevent further damage. We’ll discover how VeraViews helps train new AI models later in this article, but read on to discover how VeraViews leverages these nascent technologies.
How Does VeraViews Use Artificial Intelligence and Machine Learning?
VeraViews is an open ledger advertising ecosystem built around Verasity’s patented “Proof of View” (PoV) fraud identification technology. We utilise AI and ML in our proprietary adstack to enhance the effectiveness and efficiency of our advertising solutions.
Our AI and ML technology lets us analyse large amounts of data from past campaigns to identify patterns, trends and insights in order to optimise ad delivery in real-time, adjusting bidding and targeting based on the performance of each ad impression. This helps improve ad relevance and increase click-through rates (CTRs) which are important in today’s competitive advertising landscape.
VeraViews uses Verasity’s proprietary “Proof of View” technology, which is also powered by AI and ML, to detect and prevent ad fraud. By analysing patterns in user behaviour and identifying unusual or suspicious activity, we can filter out invalid views and ensure that our clients’ advertising budgets are being used effectively.
At VeraViews, AI and ML are at the core of our advertising solutions alongside open-ledger technology, and we are constantly pushing the boundaries of what is possible in the industry. That’s a good overview, but let’s discover exactly how the VeraViews adstack functions, and where these AI and ML modules are deployed, by exploring VeraViews’ phases of operation.
Phases of VeraViews’ Operation
VeraViews has three main stages of operations: an initial phase, a viewing phase and a scoring phase. These phases all happen when a user watches a video that is protected by VeraViews.
1 — The Initial Phase
Proof of View first attempts to recognize a viewer for a ‘pre-filtering’ score. This happens right before the ad impression is served. When a video ad is served, several steps take place in the background. Let’s take a quick look at this process.
First, a user ‘requests ‘to view a video on a website or app by navigating to an auto-play video or clicking play on a video player. The website or app then sends a request to an ad server to request an ad to be displayed to the user. The ad server receives the request and sends back information about available ads that match the criteria of the website or app, such as the ad format, length, and target audience.
The website or app will select an ad from the available options and send a request to the ad server to serve the selected ad. The ad server then sends the video ad to the website or app to be displayed to the user. This entire process takes place in milliseconds.
With PoV’s pre-filtering checks during the Initial Phase of VeraViews operation, the adstack can identify and prevent views likely to be fraudulent.
2 — The Viewing Phase
If a user or viewer passes the initial phase and is served an ad, VeraViews moves to the Viewing Phase. During the Viewing Phase, PoV gathers statistics on viewing behaviour and provides the viewer devices with proprietary challenges that develop additional data.
VeraViews is not an intrusive media player and does not challenge the viewer with a captcha like prompt. Instead, all of the challenges happen in the background without interrupting the actual ad impression or video viewing process. This is very important for ad viewability.
3 — The Scoring Phase
During the scoring phase, PoV analyses a comprehensive combination of user data that was gathered during phases 1 and 2. There are over 200 touch points in this process. Some of the data includes:
- IP address,
- OS and version,
- Preferred language,
- Cookies,
- Mouse events,
- Browser web page size,
- Volume settings,
- Timezone,
- Touch Support.
VeraViews then uses proprietary artificial intelligence to score the data which determines if the ad view is considered to be valid or fake or if the viewer behind this ad view is a real human or a malicious software/bot. This is the only known system for achieving real-time fraud detection and preventing ad spend for fake views at the source.
Data Collection and Determining Real Viewers
Once the PoV module is executed on the client device, it starts sending the data that is processed by the data normalisation module and prepares the incoming data for future processing. First, VeraViews matches a user’s unique fingerprint (using the detection data collected above) with user matching data. When combining this data with the IP address, the user matching module determines if this user has a previous history and can be found, for example through cookies (but always in accordance with GDPR).
If the ad view is predicted to be valid and the user is known, these attributes are used during the evaluation process to accurately predict future user actions. All of this happens in real time.
If the outcome appears to be higher than the set threshold (meaning the user is most likely a human) the ad is processed normally and the ad content starts playing.
Viewability parameters, content playback rate and interactions are recorded during the ad impression and optional automatic background challenges are sent to the client device to determine its dynamic characteristics, verify the content playback and more. Further data analysis determines whether the views were indeed watched by humans. This data is stored transparently, in real-time, on the VeraChain open-ledger. The publisher, advertiser and other actors in the advertising supply chain can verify and audit this data. This is, to our knowledge, the first time that such data can be verified in real-time.
Once an ad impression is complete and we’ve stored the statistics, the data is fed to VeraViews’ machine learning module where a trained AI model processes the data and provides the detection result/rating which is stored with the ad impression record on our open-ledger. These accumulated statistics, received by the machine learning module, are used to train the next generation of our AI models, ensuring flexibility and the ability to adapt to changes in the incoming invalid traffic. In other words, VeraViews not only deploys AI and ML in its adstack, but it directly contributes to the training and improvement of future AI models for preventing advertising fraud.
The PoV approach is modularized to provide analytics and filtering for datasets that differ from ad viewing stats, allowing the application of behaviour estimation and prediction using infrastructure from various data sources.
Developing New Standards for Ad Fraud Prevention
VeraViews is an emerging solution for preventing ad fraud across the video programmatic supply chain. By merging cutting-edge technologies such as AI, ML, and open-ledger ecosystems, we can help to develop new standards for ad fraud prevention.
Our tech not only deploys AI and ML in a live environment to prevent ad fraud, but it’s using the data from these live campaigns to train and improve future AI ad fraud detection and prevention technologies.
Follow Us:
- Twitter: https://twitter.com/verasitytech
- Discord: https://discord.gg/verasity
- Reddit: https://www.reddit.com/r/verasity
- Telegram Official Channel: https://t.me/VRAchannel
- Telegram Token Discussion: https://t.me/VerasityPriceChat
- Telegram Token Announcements: https://t.me/verasity
- LinkedIn: https://www.linkedin.com/company/verasity
