Results from Tag Updating
Back in early June of this year I started what has turned into a 6 month project updating all of the tags on my nearly 2900 videos, using the suggested tags available on the paid levels of Tubebuddy (TB). In late June I thought it would be a good idea to actually see if this was doing me any good. So I started to record the starting point of 5 or 6 of the 20 videos I updated each day. I did this for about a month, ending up with data on 92 videos. This was hardly a truly scientific experiment since there are far too many variables, both known and unknown. Rather than taking the results as proving a specific result, I think they point in a general direction of the effectiveness of properly tagging your videos; even if the tagging occurs over two years after the video was released.
There are many variables which I inadvertently introduced in the course of this updating. These include fixing some typos within titles and descriptions and an occasional Facebook promotion. I also added some videos to playlists created after their original release. As the experiment proceeded I became more aware of the various functions within TB tag suggestions and sorting. Since the ultimate goal was not scientific purity, but rather improving the performance of my videos, I made adjustments to my procedures as I went along.
Probably the most significant change which I began rather late in the process which may also affect these results was consistently setting the order of the tags by highest ranking first. I did this occasionally earlier on but was rather inconsistent in this step. I became more consistent later in the project. It will be one of the steps included when I revisit my tags again in a few months.
The Long Tail
With about 2900 videos, I have a very clearly laid out head and a very long tail to my channel. I was coincidentally reading Chris Anderson’s “The Long Tail” while beginning this project. It greatly helped me to understand what I was seeing as I worked through my primary playlist which contains about 2800 videos. I found a few videos which were clearly the most successful and then things tapered down from there. Anderson talks about how companies like Amazon, eBay, and YouTube have discovered the value of these very long tails which may individually not produce much traffic, but in their aggregate can produce as much as half of their profit. (For clarity: almost every business will have a few items which account for a very large portion of their sales, this is the head. The tail is the rest of their products which sell fewer and fewer items.)
Anderson describes a phenomenon called “tail thickening.” This is simply the process of increasing the visibility and sales for those products found way out on the tail. There is an outsized impact this can have on the bottom line. Anderson described exactly what I had hoped to accomplish with my project. In my youth I had worked with a secondary steel broker. He helped steel mills find buyers for their mistakes; these mistakes could often be measured in the 100,000’s of pounds. I learned the value of large numbers and how amounts as small as ½ cent could amount to significant profits. So even if I could only improve the performance of the 2500 videos in my very long tail by a small percentage, they could have a very significant impact on my channel’s overall performance.
Since the vast majority of my videos fit into the same small niche, I have used a set of pre-selected tags which get automatically loaded with each upload. In addition to these, I would add the title of the video and a couple variations to that title. Not a great plan. While I still utilized the pre-selected tags, I now spend more time looking for other possible tags.
The purpose of this experiment was to determine what, if any, benefit I could expect from updating all the tags on all of the videos on my channel. It will take between 80 and 90 hours of labor to update my tags. (I also set up end screens at this time. I had used the bulk update feature of TB to do some, but because the aspect ratio of my videos varys so much, this was not a terribly effect way to do this.)
So I proceeded to delete all of the one word tags; keeping the title and title related tags. Using TB, I uploaded all the multi-word tags which would fit within the 500 character limit. I retained a couple of one word tags which would top off the list, utilizing the last few characters. When I began my record keeping on 6/26/17, I was working on videos which had been uploaded in March of 2015, so they were approximately 115 weeks old.
With 115+/- weeks of history, I felt comfortable with just assuming at this point that each week should represent 1% of the total historical views. Yes, I know that the number is really closer to .8%, but I prefer the slightly higher bar. By the end of the test the first videos would have a history of about 130 weeks so the actually weekly number should average about .77% of the total views, but for simplicity sake, I kept the 1% threshold as a goal.
The starting view counts ranged from as low as 300 to over 5000. As a general rule, I found videos which started with a higher view count improved the most. Although, there were a couple who didn’t improve much at all, they usually had some kind of high view event in their history which inflated their view counts. (One of the variables I didn’t control for.) (I also had one video with significant errors in it which I had delisted. I didn’t catch this immediately and left it in the program even though its growth was close to 0%)
So here are the results at various time intervals. The % gain is per week. The base gives each week’s portion of the historical average.
|Approx. channel growth**
*This is the breakdown of the weekly average for the immediately preceding 4 or 8 weeks.
**Approximate month/month view growth channel wide. Some of the growth could be attributed to being part of a growing channel, but not all.
I think this provides some pretty good evidence of the benefits of updating your tags periodically. My channel experienced a steady acceleration of view count over the course of this project. Also, I have taken other samples, and while these experiments are still running, they seem to confirm these findings. I found improving tags with the help of TubeBuddy’s suggestions improved performance by about 40% per week over a channel wide baseline. While in many instances the raw view count was still pretty small and may not sound like a lot. However, small numbers, multiplied by large numbers, over a long period of time generates huge returns over the long term.