An Easy YouTube Framework: For Testing Keyword Success

So “content is king”. But seriously, can you say that a video with a million views is always better content than another video with only a thousand views? Even more puzzling is that videos with more views don’t always rank higher in YouTube Search Engine Rankings.

We wager that in 2013, SEO practitioners are increasingly going to face questions about what drives ranking in YouTube. A great place to start is this Whiteboard Friday from Aug-12.

The impact on Search Engine Rankings (SER) from better selection and placement of keywords in meta-title are well-researched topics. Now we all know that search results on YouTube work somewhat differently. See image below comparing top four results for “how to” in Google Video and YouTube.

So, given the growing importance of YouTube, we have developed an easy to execute framework for keyword analysis. You could use this framework while deciding between synonyms, or while assessing the benchmark views-per-day, or just to get a better perspective of what’s driving views and ranking on YouTube.

Our assumption while developing this framework is that while uploading a video on YouTube, your objective is to maximize views. In our experience, most objectives do have a strong correlation with views, e.g. number of external embeds, viewer comments. But to be fair, there may be that odd case where your objective may not necessarily be correlated with maximizing views. That said, let’s get going.

In this post we use the framework to present some observations while comparing YouTube videos with “How to” in their title and “Tutorial” in their title. The questions we are seeking to answer are,

• Which keyword is more likely to get higher Views Per Day
• Do older videos get any noticeable preference in YouTube SER
• Does the number of subscribers influence the Views Per Day
• Does the location of the keyword impact YouTube SER

It’s well understood that YouTube is the home for “How to” videos. “How to” videos are so important to YouTube that on 11-Dec-12, YouTube announced the search for the next “How-to Guru”. YouTube search currently shows over 131m videos having “How to” in their title. “Tutorial” is only a fourth as popular with 35.3m videos.

If you believe that 131m videos can’t be wrong, our findings will surprise you! But before that, here’s some more ammo. There are 10x more searches on Google with “How to” than “Tutorial”.

For the purpose of this analysis, we looked at the first 25 pages of results (on YouTube) for “How to” and then for “Tutorial”. So for each keyword we had a sample set of 500 videos to start with. We then went to each video page to tabulate data for our analysis. For more details, see the D-I-Y section below.

1) Videos in the “Tutorial” sample are doing 20% more views per day! This being in spite of the huge amount of search traffic directed towards “How to”. Clearly, the supply and demand is bringing an unexpected balance (ps: an additional point for those who are new to YouTube analysis, 174 views per day is not at all a bad number).

Table 1: Views Per Day

2) At first we were excited to find that the average age in the “How to” sample was 41% more. But on closer examination we found a low correlation between age and YouTube SER. So we’re really not ranting about this.

Table 2: Age and YouTube SER

3) Here’s a really interesting one from the correlation between Views per Day and Number of Channel Subscribers. The potential takeaway being that if you are a young channel, you’re probably better off using “Tutorial” as a keyword, and building up your subscriber base, before considering switching to “How to”.

Table 3: Views Per Day and Channel Subscribers

4) It appears that there is no advantage to be derived from the placement of “Tutorial” earlier in the video title. However, there does appear to be some correlation in placing “How to” upfront in the video title. But here we are hesitant to draw causation, given that popular grammar usage mostly leads to sentences starting with “How to”.

Table 4: YouTube SER and Keyword Location

Blindly taken, our findings above would imply that for most cases, we should ditch “How to” in favour of “Tutorial” (as a keyword in You Tube). But this is a framework for analysis. There is the additional step of human intervention before making a decision. In general, if the framework throws up a finding that goes counter to your experience or intuition, we’d strongly recommend discussing your findings with a colleague or friend.

We started this post by saying that rankings on YouTube are puzzling. In our study, here are two videos using “How to” that puzzled us:

Good video with low views per day (Video A): Adobe InDesign CS6 : InDesign CS6 | 3. How to Create a Promo Flyer [for Beginners] doing c. 76 views per day (not bad per se).

Average video with high views per day (Video B): How To Correctly Serve a Watermelon doing c. 11.9k views per day!

Now with the benefit of the above analysis, we can say that if Video A was being uploaded again, using the keyword “Tutorial” would probably be a better strategy. That said, there’s no way of ensuring that Video A would beat Video B in a re-run.

D-I-Y (More power to you!)

Maybe understanding “Football” vs. “Soccer” is more important for you. If so, conducting the above analysis is really not that difficult. Here’s an Excel Template that should get you going. We’ve inserted comments that should help you navigate and run your own analysis. However, if enough people want, we can put up a YouTube tutorial.

If all you need is a quick analysis, we suggest randomly picking up 35 videos across 25 pages – for each of your keywords. That should be a statistically representative sample for analysis.

What follows below are Notes to answer some of the questions that may be on your mind. Lastly, Tweet @toytasting and we will send you a direct link to download both the Excel Template and a PDF of this article.

What’s the weakest assumption in our narration? (Big Geek High Five here!)

We didn’t elaborate above, but if you have reached this far, you’ve probably caught onto it already. You see, we’ve placed some faith on the metric “Views Per Day”. Now the plus of this metric is that it normalizes the speed of views, for old and new videos, and makes it possible to compare performance. The catch is that views on YouTube are patently non-linear.

A more thorough analysis could be done with measuring speed as well as acceleration.

How about a deep dive regression, instead of just correlation?(Super-Big Geek High Five here!)

Yes, we did test regression with straight-line, exponential and power curves. But we were not happy with the R-square that we were getting.

Could outliers affect the conclusions? (Geek Ecstasy!)

Not really. As we’ve used median instead of average, as a measure of central tendency. However, that can’t be done for computing correlation. Still in our experience outliers would impact attempts to curve-fitting (see above paragraph on regression) more than a pure correlation computation. So, we think that the conclusions from this framework are not fragile to outliers.

What is correlation?

Correlation is a statistical computation to measure the extent of linear association between two variables. Correlation is always a number between +1 and -1. For everyday use you only need remember the following,

• If the association is positive the two variable move in tandem (e.g. Celsius and Fahrenheit) and when negative the two variable move in opposite directions (e.g. altitude and temperature),
• +1 or -1 denotes lock-step correlation or perfect correlation. 0 denotes no correlation,
• From our experience, we discard correlation in the 0.5 to -0.5 zone as not being material

In MS-Excel the formula to use is CORREL(array1,array2).

Here’s a great page explaining correlation.

How to check the number of videos using a particular keyword in the title?

Here’s an example of what to type in the Google Search Box, if you wanted to see the list of videos on YouTube, using the keywords “how to” in their title > allintitle:how to site:youtube.com

Google Trends

We’re big fans of Google Trends (isn’t everybody!). Don’t consider any analysis complete without at least spending some time on Google Trends. Here are a few interesting facts that came up in the context of “How to” vs. “Tutorial”. We believe these insights can be leveraged in interesting ways – but that’s a topic for another blog post.

• The top-3 cities from which searches for “how to” are made: Bangalore, Singapore, Los Angeles
• The top-3 related terms while searching for “how to”: google, how to love, how to draw
• The top-3 cities from which searches for “tutorial” are made: Bangalore, Chennai, Jakarta
• The top-3 related terms while searching for “how to”: tutorial photoshop, photoshop, java

Richa Dikshit runs the YouTube channel of Toy Tasting – a video channel about toys.

Vishal Somaiya works at Yellow Giraffe – an online toy store in India, where he leads inbound marketing.