If you pay attention to your view counts for videos on a YouTube channel you own, you might expect them to be relatively stable. After all, it doesn’t seem like there can be a “negative view”, so the number should never go down. Yet, by monitoring a video for a few weeks, you can often clearly see view counts go up and down, spiking and dropping, often unpredictably. What causes these drops?
The answer is that YouTube, and Google in the background, has a lot of quality control measures in place. However, in order to maintain a smoother view count experience, YouTube has decided to perform auditing and validation on a rolling, ongoing basis. This is in contrast to their old method.
The Old Way
In years past, YouTube would stop any video that reached over 300 view views and halt the view count display. This was the “301+ error” that so many people used to encounter. You would see a video, and the view count would be listed as 301+, rather than an accurate number. Sometimes it would even sit at this number for days or weeks before it changed back to a real number and started tracking again. While frozen, you would be unable to see a real, accurate view count.
Why this freeze? The answer was the YouTube view quality audit. Essentially, Google didn’t really care about videos with under 300 views. If you wanted to fake 140 views for your video, you could do so, and they wouldn’t even notice. There are simply too many people uploading too many videos that barely get any attention at all. There wasn’t enough processing power to audit all of them from the beginning, so YouTube didn’t bother to try.
What they did is set a threshold, that being 300 views, after which they would trigger an audit. The view count would freeze at 301+, and they would set their algorithms loose on it. They would analyze records and behavior of both old views that were recorded, and new views coming in while they audited everything.
YouTube would be looking for signs that the views were fake in some way. There are a lot of possible sources of fake views, so they had a lot to analyze.
- Do a lot of the views last under 10 seconds? A lot of fake views simply load the video and then leave.
- Do a lot of the views come from the same IP address? It’s unlikely that the same person is watching the same video over and over. This can be legitimate behavior – if you like a song you might watch the video several times – but for less compelling or less repeatable content, it’s a lot less likely.
- Do a lot of the views come from unusual or duplicate user agent strings? Sure, there are people running around using old versions of Safari or Internet Explorer, but not so many that it’s going to make up a significant portion of your view count. If a lot of views come from strange user agents, it’s a sign that a lot of those views might be bots.
- Do a lot of the views come from the same referral URL? This can be natural, such as if the only link leading to the video is from your blog and it got a lot of traffic. However, it can also be unnatural, like in cases where the link is embedded in an advertisement rendering the video. Paid traffic isn’t technically against YouTube’s rules, but they have to know about it, so using shady paid services can be a bad idea.
The end result is that, when you hit 300 views, it will take a short time to process and will revert to a “real” view count. This could be higher, if you continued to get legitimate views during that time, or it could be lower, as a bunch of fake views are removed.
Some views will always be removed. There are bots that do nothing but trawl for links around the web. They’ll click through to videos and “watch” them for a second, before moving on. There are also bots that simply crawl YouTube to add views to whatever they come across. YouTube filters a lot of these views, but it often takes the audit for them to be totally removed.
The New Way
YouTube was tired of answering questions about the 301 view pause, and they were tired of having to deal with the issues it caused. It’s not natural looking anyway, and it often seemed to stunt growth on videos right as they started to pick up. Seeing no progress and encountering an audit wall made content creators disheartened about further investment in progress they couldn’t see. It’s sensible that YouTube changed how they process views. Plus, as computers have improved over the last decade, the YouTube data processing centers have grown more capable of auditing and the auditing algorithms have improved dramatically.
So what happens with a modern video? Well, first of all, there’s no audit threshold, because audits begin immediately.
Many modern videos seem to start off slow, because YouTube is very conservative with their audits. They do a first pass audit on every view a video gets, looking for every possible sign that the view is less than legitimate. It’s very strict, and only the views that are 99% likely to be real are let through. Any view you get right away is guaranteed to be a real view, or the sneakiest of sneaky bots sufficient to obfuscate itself through YouTube’s filters. Mostly, it’s real views.
After that, YouTube starts doing more detailed but less harsh view audits on the views your video has gotten. They soften the requirements overall, but test specific elements more stringently, to filter out views they know are bad and let through views that are moderately likely to be good. This process happens on an ongoing basis, so every new view you get is audited to see if it’s real.
The end result is that your view count will start off by rising right away, and then will stagger up and down as the audits pass over it. Some views will be added, some will be taken away, and in aggregate your view count will rise and fall overall.
One thing to note here is that you can use YouTube Analytics, which is a branch of Google Analytics natively available to content creators on YouTube, to see a more accurate version of the view counter. The public view counter is mostly just an estimate. However, the one piece of data you probably want to reference is the real time report for views, and that is itself just an estimate. It is subject to the rolling audit as well. It will show you the views you’re getting in real time, but those views include the fake views that will be removed from the final total.
Another thing worth mentioning is that views do not equal earnings. A video with 1,000 views can earn more than a video with 10,000 views, for a number of reasons.
The first reason is that different niches will pay different amounts. Just like with PPC advertising, different niches can have wildly different pay rates. That’s why you see all of those class action settlement websites and pharmaceutical sellers; they have high commissions, even if there’s a lot of competition. It just comes down to economics. There are a lot of possible buyers for a cheap item in Gaming, conversion rates are low, and profits are low, so the earnings for advertising on a gaming video are low. Meanwhile a health video on a specific narrow topic might have an almost nonexistent audience – only a few tens of thousands of people in the world could be affected by the issue, and only a fraction of them will see the video, and only a fraction of them will convert. Each conversion is worth a lot, though, so advertisers are willing to pay for it.
The other reason is that YouTube today works on a view time model more than a view count model. Advertisers understandably don’t want to pay for three minutes of advertising when the user only watches five seconds of it. You tend to earn more on a video where people watch it all the way through, rather than a video where people drop it quickly. Viewing time has become an increasingly important metric.
The Great Data Error
A year or so ago, YouTube had a massive data reporting error. The glitch, as support staff called it, caused a lot of extra views to show up on videos all over the site. It took about 36 hours for this glitch to be fixed, and when it did, view counts went back to normal.
The problem with this came from people who watched their real time view counts or just monitored their video views on a daily basis. These people saw their videos doing unexpectedly well, and then suddenly saw anywhere from 200 views to 50,000 views simply disappearing. I have no doubt that it happened to larger video producers as well, but most of them are smart enough to read around rather than post in the Google support forums in a huff.
The fact is, a lot of people were complaining about their views disappearing, when in reality there were a lot of views being reported that never existed. The number was wrong and it was corrected. To people who don’t have access to all of the data, though, it looks like views they earned disappeared. It was very hard for some people to understand.
There was also a similarly devastating-looking glitch that happened as recently as earlier this year, back in January. What happened was that a glitch in the code that monitored subscriber counts caused an unsubscribe to count for more than one subscriber. If you had 100 subscribers and someone new subscribed to you, it would go to 101. If that person then unsubscribed from your channel, it would drop you by 2, to bring you down to 99. You still had 100 subscribers, but the display glitch made it look like a person could tank the subscriber counts of any account they wanted.
This was abused for several high profile accounts, and people were even building bots to rapid-fire subscribe and unsubscribe from various channels they didn’t like. Top channels with controversial content, like PewDiePie, were hit for millions of unsubscribes.
It was essentially just a fun day as far as most people were concerned. Google stated that it was a display bug and that the YouTubers weren’t actually losing their subscribers, so the subscribers were able to relax and wait for Google to restore their counts. Everyone else would either be banned for botting or just ignored. It made for a few funny news stories, and not a whole lot else.
The point of me bringing all of this up is to set the stage for the following statement: YouTube is not perfect. I know you know this, and I know it’s not hard to picture. Essentially, at any time, another glitch might happen. If you’re monitoring your view counts and you see a sudden drop, it could be an audit, or it could be a data management glitch on YouTube’s end. If the drop is unusually significant, it’s entirely possible that YouTube was breaking something and they fixed it, or they’re breaking something presently and will fix it in the future.
It’s also always possible that some third party has decided to embark on a campaign of negative video SEO, buying views for your videos in an attempt to get the mass bot views audited and your videos removed for buying fake views to them. This happens, though it tends to be rare and tends to require quite a large view buying investment. I wouldn’t worry too much about it.