*This article/analysis is not politically motivated
Big Tech bias, is it real? What can OpenRTB tell us about the accuracy and relevance of YouTube search results?
Big Tech - Subjective search results favoring left leaning content - Reflection on OpenRTB
For many years now we are hearing complaints coming from ‘the right’ regarding supposed bias and censorship from big tech. This supposed phenomenon is being experienced on different social media platform and search engines. More specifically, do Google and YouTube favor center and liberal/left leaning content in their search results?
In this article we focus on the accusations of the Trump administration concerning censorship and bias by Big Tech. We will reveal results of our internal investigation regarding apparent subjective search results in YouTube, favoring center and left/liberal content - and how we deployed a ‘thermometer’ in the digital ecosystem of display advertising (OpenRTB) to determine if there is any validity in this phenomenon (by this time possibly experienced/noticed by hundreds of millions of YouTube users - if not more).
Why cant’ display advertising on OpenRTB be controlled by Big Tech?
OpenRTB, in many ways, can’t be censored or controlled by Big Tech. Simply put; the methodology behind search results (either in Google or YouTube) are so called ‘black box’ software. Meaning; we, the people, but also government - have little to no idea how the search result algorithm actually works. Hence, YouTube is in complete control of their search results (the same goes for Google). The difference between this nontransparent black box environment (YouTube) and transparent OpenRTB is that - in the case of OpenRTB and by using a strong DSP - ads can be served on every website and application that allows ads (or; has adslots) - without approval from Google/YouTube. So again; a DSP that is integrated with many adexchanges (which in turn have thousands of apps and sites connected to it - for monetization purposes), can serve your ads on almost all apps and sites that allow advertisement. Although Google has it’s own adexchange, it’s just one of 64 adexchanges/supply sources we are integrated with. If we decide to run ads on publishers that use the Google adexchange, Google has to approve the ad in question. If we want to serve ads on publishers (sites/apps) that are not bought at Googles’ Adexchange, we DON’T need Googles’ approval to serve the ad. And here is the thing; the total amount of publishers (again; site and apps) that are integrated with the Google Adexchange represent about 8% of total apps and sites that allow advertisement. Not to mention the fact many publishers which can be reached on Googles’ Adexchange, are also - duplicately - available on other adexchanges.
YouTube search results
If you search for ‘Trump’ in the YouTube search bar, usually the first 5 to 9 results can be described as negative. Meaning; the actual ‘news’ (in this case; a video) that each results contains is not positive about Trump. It’s also very noticeable that the far majority of search results (again; when searching for ‘Trump’) is content from left leaning news outlets. It doesn’t matter from which country or from which device you are entering ‘Trump’ in YouTubes’ search bar, the results consistently stay the same; far majority of search results are videos with a negative tendency towards Trump. Due to internal worries about infringement etc., we have decided to not publish screenshots of these YouTube search results. However, we challenge you to do your own research by simply entering ‘Trump’ in the YouTube search bar and observe the results.
Ethically speaking you could say that any search engine should show results based on popularity and relevance. When this is not the case, it’s possible that a – by origin - commercially motivated organization is interfering with and decides on; reality. In this case; ‘reality’ in the form of search results and ‘trends’.
In the recent years it has become clear that one political movement specifically feels they are undermined (understatement) by Big Tech. This is of course the conservative political movement (when talking about the USA). Many research and published content suggests that this undermining of the conservative movement is real.
Our ‘Thermometer’
Back to us (Targetoo) and OpenRTB. Is there a way to shine a light on this supposed bias? More specifically, being that OpenRTB is minimally controlled by Big Tech, is there some sort of experiment we can do to prove or disprove there is bias present in search results?
What did we do:
We established an unusually small whitelist consisting of 3 right leaning publishers (sites/app) and 3 left leaning publishers. As follows:
Right:
Breitbart
ZeroHedge
RT
Left:
CNN
HuffingtonPost
NBC
On each site we served a ‘pro-Trump’ and an ‘anti-Trump’ ad (note the low quality banners. No advanced Rich Media ads). As follows:
300x250 ‘Negative’
300x250 ‘Positive’
The results of the ‘anti-Trump’ banner:
Publisher (center/left leaning):
CNN 1250 imps 2 clicks 0,16% CTR
Huffington Post 1250 imps 1 click 0,08% CTR
NBC 1250 imps 0 clicks 0,00% CTR
Publisher (right leaning):
Breitbart 1250 imps 0 clicks 0,00% CTR
Zero Hedge 1250 imps 4 clicks 0,32% CTR
RT 1250 imps 2 clicks 0,16% CTR
combined CTR of ‘negative’ banner = 0,12% CTR
The results of the ‘pro-Trump’ banner:
Publisher (center/left leaning):
CNN 1250 imps 9 clicks 0,72% CTR
Huffington Post 1250 imps 17 clicks 1,36% CTR
NBC 1250 imps 10 clicks 0,8% CTR
Publisher (right leaning):
Breitbart 1250 imps 22 clicks 1,76% CTR
Zero Hedge 1250 imps 23 clicks 1,84% CTR
RT 1250 imps 16 clicks 1,28%
combined CTR of ‘positive’ banner = 1,29% CTR
Reflection:
Admitting, we were expecting to see the ‘negative’ banner perform better on the 3 left leaning publishers. Likewise we were expecting the ‘positive’ banner to perform better on the 3 right leaning publishers. In both scenario’s, this is not the case. Although the average CTR differs slightly, the difference is to little to reach a clear conclusion and/or see a tendency.
The ‘positive’ banner, regardless of being served on either 3 left leaning publishers, or 3 right leaning publishers - has a substantial higher Click Through Rate (1,29% versus 0,12%) compared to the ‘negative’ banner.
The ‘negative’ banner, regardless of being served on either 3 left leaning publishers, or 3 right leaning publishers - has a substantial lower Click Through Rate ((0,12% versus 1,29%) compared to the ‘positive’ banner.
The most profound conclusion there is to be determined, is the fact that the ‘positive’ banner (positive news regarding Trump) performs exponentially (975%+) better than the ‘negative’ banner. Regardless if the ad is served on either left leaning publishers, or right leaning ones.
One might say we are comparing apples (YouTube search results) and peers (clicks on a banner). However, clicks are a solid metric to determine popularity of a certain subject. Whatever the motivation behind the substantially larger click amounts on the pro-Trump banner is (frustration or joy and possibly everything in between), the results speak for themselves. The overwhelmingly negative search results when searching for ‘Trump’ in YouTube - do not reflect OpenRTB reality.
Do you want to discuss this sensitive analysis. Reach out. Likewise, request the full analysis here.