With Juniper Research estimating advertisers will lose $42 billion of ad spend globally in 2019 to fraudulent activities -- up 21% from the prior year -- marketers need a way to fight back. By 2023, that number is forecast to reach $100 billion.
Steven Ohrnstein, SVP of platform automation and analytics at Viant, said the Irvine, California-based company has been working for years to wipe out fraudulent impressions from bid requests with a set of new tools.
“We found that about 10% of traffic is potentially fraudulent ad inventory that we don’t allow brands to bid on,” Ohrnstein said.
Viant’s technology runs on Google Cloud services and enables the company to process tens of billions of ad requests daily, each with about 700 data points, Ohrnstein said.
“It enables us to crunch every possible permutation with a machine learning algorithm,” he said. ‘Without a cloud service this would be impossible.”
The new fraud prediction tool uses machine learning to understand past performance and the likelihood of the bid request being fraudulent. A score is then created. If the request is at all suspicious, it is automatically discarded from the bid request.
The viewability prediction score also uses machine learning, but this time to identify highly viewable inventory and how much to bid for an individual impression. This feature also creates a score to determine the likelihood of being in view.
The bid optimization tool helps to lower the cost of the media per campaign. Agencies typically want to lower the budget based on the key performance indicators to help the brand, but it’s typically not in the best interest of the agency, unless that agency can use the funds in another campaign.
The agency will typically submit a maximum bid in with any DSP, and that is the bid paid on every impression.
Viant’s machine-learning bid optimization tool focuses on automating bids and taking the bid out of the trader’s hands to drive down the minimal price while following the KPIs based on historical data.