Study: ‘Beware of CTV ad fraud’
July 16, 2020
Cross-screen advertising technology specialist Sabio has revealed the results of an in-depth, connected TV (CTV) fraud analysis from its App Science technology platform. For the study, App Science cross-referenced location, census, traffic, content, device technographics, and third-party app data with more than 150 million IP addresses to identify some of the tell-tale signs of fraud in the CTV space.
“CTV fraud schemes like ICEBUCKET, DiCaprio and Monarch have already siphoned billions of dollars from advertisers,” noted Sudha Reddy, VP of Product Innovation at App Science. “The industry must learn from these operations, identify the tactics behind them, and proactively put measures in place to weed out this suspicious behaviour in the future.”
While fraudulent methods like server-side ad insertion (SSAI), platform mismatch, and app ID spoofing have been behind the high profile CTV fraud schemes, there are a variety of indicators that can serve as early warning signs. Some of the suspicious behaviour Sabio’s App Science team identified include:
- Improper Site IDs: only 12.5 per cent of Roku site IDs meet the IAB standards. The remaining 87.5 per cent have their own naming convention, which is not necessarily fraud but opens up risks when buying inventory.
- ‘One-Time’ Devices: a large portion of devices (22.4 per cent) are classified as ‘one-time’, meaning they have device IDs that have been seen only once in the system. This is suspicious and needs to be investigated further, since people typically use the same connected TV or streaming stick to view video content repeatedly.
- Operating Systems: 5 per cent of CTV devices and operating systems are obsolete, which should be monitored to ensure that traffic coming from these sources is declining. When traffic from these sources increases in certain weeks, this needs to be evaluated further.
In addition to these fraud indicators, Sabio suggests it is also important to audit media content which can correspond with spikes in traffic and viewing patterns for different apps. For example, data shows that news content is streamed more in the afternoon (peaks at 14:00), while movie apps are used later in the evening (peaks at 03:00). These are all things that must be monitored, especially with Upfront spending down 33 per cent this year and advertisers turning to programmatic to allocate an exponentially larger portion of their marketing budget.