OTT — or Over-The-Top content — has been used by the media industry to refer to the access of video content over the internet via a broadband connection. These connected devices are most typically gaming consoles, smart TVs, streaming boxes, BluRay players and similar devices.
However, according to this definition, smartphones or laptops could also be interpreted as OTT devices. Furthermore, with these connected devices, a user may view video content that either directly replaces the use of the direct medium — such as watching HBO content in a web browser or within the HBO Go app instead of on their available HBO channel — or they may consume content ordinarily unavailable via their content provider, such as YouTube videos.
As cord-cutting becomes the norm, the definition of OTT becomes blurred. For example:
In order to eliminate any confusion, Pixalate has deployed a multi-dimensional approach within an application-appropriate taxonomy to define the ways to connect to media and the nature of the content itself.
Once we have derived platform and media type from a combination of known user agents and our own script data, measurements are conducted in a similar fashion across all devices and traffic is classified as viewable, not viewable, or not measurable. And where measurable, traffic is classified as GIVT, SIVT, or valid.
Pixalate breaks down fraudulent behavior into many distinct types. Below are some — but not all — of our major fraud types:
Fraud Type |
Definition |
|
Auto Reloader |
Impressions with very periodic pattern that cannot be generated by a human. |
|
Cookie Click Fraud |
Clicks that are generated from the same cookie at a rate of more than 1 per minute. |
|
Cookie lmpression Fraud |
Impressions that are generated from the same cookie at a rate of more than 1 per second. |
|
Data-center |
Traffic that is coming from IPs that belong to data-centers. |
|
Device ID Click Fraud |
Clicks that are generated from the same mobile Device ID at a rate of more than 1 per minute. |
|
Device ID lmpression Fraud |
Impressions that are generated from the same mobile Device ID at a rate of more than 1 per second. |
|
Fast Clicker |
Clicks that occur in less than 1 second apart from their respective impression. |
|
IAB Crawler |
Software robots that use a User Agent string that does not belong to any existing browser. |
|
IAB Dummy Bot |
Software robots that periodically visit websites to crawl their content while revealing their identity in their User Agent string. |
|
Idiobots |
Bots (or users) that change their User Agent string (spoofing), while keeping the same cookie. |
|
Malware |
Domains or pages known to host malware |
|
Phishing |
Domains or pages associated with phishing tactics |
|
Proxy |
Traffic that is coming from an anonymous proxy server or a TOR network node. |
|
Publisher Fraud |
Domains that use stacked or popup ads (usually after malware infection) to excessively increase the impression counters. |
|
Smart Bot |
Bots (or users) that change their browser agent string (spoofing) and cookies very often under the same IP, creating low volume traffic or high volume traffic under configuration that looks like a busy enterprise network. |
These fraud types are measurable across all platforms, including Connected TV/OTT. However, given that each platform and media combination has unique characteristics, constraints, and usage patterns, Pixalate is able to derive a highly predictive model for fraud which is also unique to each platform.
Therefore, based on this predictive, machine-learning algorithm, the classification into one of these fraud types may be more or less sensitive to the measured platform. While we can’t go into specific details about the algorithms, it is self-evident that for most types of Connected TV/OTT devices, the video medium is more prevalent than display.
Such ground truths can be employed in the measurement of events that ultimately lead to the classification of any given ad impression as viewable or not, measurable or not, or fraudulent or not.
