Introduction

This is part 2 of my lessons learned as a Tableau Blogger. Part 1, which is a qualitative examination of Tableau blogging, can be read by clicking this link. In part 1, I promised to also give a quantitative analysis of what I have learned from blogging about Tableau. Don’t worry, however, I will not do the following:

1. use multiple regression analysis on my blog history;
2. perform a cluster analysis using the blog post subcategories I have written about;
3. show you how I attempt to use finite-element solution methods within Tableau to solve second-order, elliptic partial differential equations such as the Laplace Equation.

I’ll keep the analysis simpler than those things (maybe I’ll save those topics for another day). What I will show you should be enlightening, if I do the job correctly.

---

If you like this article and would like to see more of what I write, please subscribe to my blog by taking 5 seconds to enter your email address below. It is free and it motivates me to continue writing, so thanks!

---

Background

I have a confession to make and some background explanations to offer before I unload this material on the few daring individuals that are crazy enough to want read it. First, I confess that I have never used the Tableau Superstore Sales file, or whatever it is called in all the training videos produced by Tableau. I know, I know, what you are thinking. How can that be the case? How can a guy like this, who professes to be a “Tableau Blogger”, never have used the super store file? Well, basically it happened in a particular way and there is a reason I have to tell you this, so please be patient.

Back when I started using Tableau in early 2008, the only training materials I could find were the Tableau knowledge base articles and the training videos that were available at that time. There really weren’t too many Tableau bloggers at that time, so since I was using Tableau on real-world consulting gigs, I basically taught myself how to do things while feeling the pressure to produce project results. Occasionally I’d send a note to some guy named Joe Mako and ask him if he knew how to do something, but for the most part I just figured things out. If I couldn’t figure it out, I’d consult the knowledge base articles (I even built a Tableau database to make these searches easier). If I still couldn’t figure it out, I’d call Tableau and ask for help. I had to do that a couple of times in the early days of learning Tableau but haven’t had to do that for a very long time. I have continued to learn on my own and by reading a number of emerging (and outstanding) Tableau-focused blogs. For the times that I couldn’t get Tableau to do what I wanted it to do, I invented my own methods.

So instead of using the super store file, I just tried to reproduce what I saw in the videos and knowledge base articles by using my own files and my clients’ files. Since I have worked with so many different clients through the years, this dramatically built my experience base by forcing me to use many different data files. This also helped me become really good at handling any type of data thrown at me by clients (and there are plenty of stories left untold from these experiences). Therefore, in my blog you will never see a superstore example (at least there isn’t one in the next 50 posts I have planned!). What you will see, however, is how to solve real-world, business type problems using a wide variety of Tableau (and other software) techniques.

Lastly, I didn’t try to rename my blog to be consistent with a Tableau theme. I have used the term 3danim8 for many year before using Tableau because I used to specialize in creating three- and four-dimensional animations of scientific data that was generated from numerical models. 3danim8 simply is a shortened version of “three-dimensional animations”.  By not using a Tableau inspired blog name, I intentionally tried to fly under the radar, so to speak, and the reason for this is given below.

 Assumptions of The Tableau Blogging Experiment

I now must make a few statements regarding the methods I used during my Tableau blogging experiment.

First, I have not tried to broadcast my blog in any other way than a single twitter post (@3danim8) at the time each blog post is written. I do not broadcast via Facebook, Google +, linked-in or any other social media. I intentionally kept my method simple to keep my analysis simple. Therefore, all posts have been broadcast the same way. I did this because I wanted to see if the old adage of “build it and they will come” would apply to a technical blog. I also wanted to learn most directly which type of content leads to the best reader response rates, so I didn’t want the effectiveness of different social media platforms to confound my analysis. The only exception to this method is my recent Tableau/Alteryx blog post where I used NFL financial information with demographic information to gain some insights into that business. In this example, I tried to use Twitter to increase the response rates for that article by appealing to NFL fans with a couple of tweets.

Second, in some of my posts I used training videos that show how the work is conducted. I started the first few posts that had youtube video links and these may have gone out to Google + back then, too, but I fairly quickly standardized my approach by upgrading my blog to a paid version to be able to place all my content in the WordPress content management system. I find it is just easier to keep things together in one place. I do not believe that this change affects the results described below at all because my first few posts were essentially never really received by the Tableau community. Their readership histories are so small as to be negligible.

Third, I did not set-up a structured experiment like I would normally do for this type of test. Normally I would execute a multi-variable test (MVT) and quantitatively analyze the results. However, at the time I started the blogging experiment I was too lazy to conceptualize the experiment in enough detail to complete an experimental design. I felt at the time that I could just get away with “winging” it. In retrospect, having a formal experimental design would have been awesome. Maybe I’ll try one just for fun!

Last, I published the posts whenever they were finished, whether it be on a Friday night at 2 am or at lunch during a workday. I did not try to investigate how readership changed depending upon when a post is published because I don’t think that it matters that much. I have found that my blog has gone world-wide in a short amount of time and therefore, someone is always reading the content.  I have concluded that the time of day a post gets published does not matter.

Quantitative Analysis of Blog-Post Response

I now start the quantitative analysis by giving two examples to examine performance characteristics of my blog over time. To do so, I need to introduce a metric that I have created called the readership response rate (RRR). The RRR is simply how many people read your blog per day (or an individual blog post per day). The RRR is cast in terms of how many people per day read my blog per 1,000 followers the primary subject has on Twitter. The first example I provide compares two blog posts to see how the RRR’s compare for two different topics, with two different communities. The second example, examines the overall growth of my blog over time.

This analysis is not exhaustive, so I will leave the possibility open for a part 3 of this series. One of the lessons that I have learned is that you should strive to be concise in your writing and to not make the posts too long, and this one is longer than I thought it would be (but I have a lot more to say).

Example 1: A Tableau-Focused Blog Post Vs a Non-Tableau Focused Blog Post

The RRR you receive from writing a new blog post can vary wildly. You can write something that goes completely unnoticed, or you can write something that strikes a chord with people.  Let’s look an an example of a Tableau-focused blog post that I wrote that generated more interest than I expected.

Blog Post A: A really good responding Tableau-Focused blog post

This post was titled “How #Tableau Has Taken Me Into the Data Zone“. At first glance, this title is completely obscure and my expectations for this post were essentially nil. I really didn’t expect anyone to read it because nobody could possibly have known what I meant by the term “Data Zone”.

As an aside, I have now come to realize that what I was describing in this article, is a real phenomenon for a lot of people. I have coined the term TDZ (or Tableau Data Zone), to describe the eight hour working days that pass by for me in minutes. On those days, my biological clock seems to get altered such that an hour seems like a minute when I am deep into my work using Tableau Software.  The concentration level I achieve is the reason this happens. This happens to me frequently and this is what I mean by working in the Tableau Data Zone.  If you want to learn more about what I mean, click here.

Now getting back to the analysis. After publishing this obscure blog post, I was very surprised by the RRR it received. Figure 1 shows the number of views for that post since it was published on Friday, May 30, 2014.

You can see that there is an initial burst of readership, followed by a second burst a few days later (due to a Tableau Software re-post of this article their blog), followed by an exponential decay of the readership. The long-term stats are 900 views over 107 days, or an average of 8.4 views per day. The peak readership in a day was 204 views on Tuesday 6/3/14, or almost 23% of the lifetime views occurring on that day. The post has now settled down to a modest one or two reads per day on average.

What does this teach us about a “hot” blog post topic in the Tableau audience? A few things can be learned. First, a “hot” blog post does not remain “hot” for long.If you want to build your readership, you have to work at it by continuing to create new content. Even well-received Tableau blog posts will only boost your readership stats for a few months after publication. You will always have a residual return from a post but do not expect people to keep coming back to read any particular article. Articles tend to fade off into the sunset just like everything else in life. “Hot” topics seem to fade in an exponentially diminishing manner. I’ll show other examples in this series that do not behave this way and have steady or increasing readership over time.

This “hot” topic insight is one reason why it is challenging to write a successful blog. Blogging require a continuous influx of ideas, creativity, time and perserverance. A growing blog requires a mix of “hot” topics and slow and steady accumulators. Most importantly, a good blog also requires an understanding spouse.

However, to continue this analysis, additional insight can be gained from this blog post by comparing its RRR to another another blog post on a different topic that is not Tableau-focused.  This analysis will help us understand how the Tableau community is receiving our attempts to communicate our insights through the blogging methodology.

Example B: A really good responding non Tableau-Focused blog post

A couple of weeks ago, I wrote a post about an ultra-endurance athlete named Rich Roll. The blog post topic was why I think he is a great podcast host. A single tweet went out when I published it, but a barrage of readership traffic followed as shown in Figure 2. The initial stats are 1240 views over 10 days, or an average of 124 views per day. The peak readership in a day was 595 views on Sat 9/6/14, or almost 48% of the lifetime views occurring on that day. The post has now settled down to an average of 6 reads per day.

To compare these two blog posts is reasonable. They both happened to be posted on a Friday (pure coincidence). The number of Twitter followers for @Tableau is 38.3K and the number followers for @RichRoll is at 31.9K. Both @Tableau and @RichRoll retweeted my initial post. Rich Roll also put a link on his Facebook page and I don’t know if Tableau Software did the same.

What I do know is that a lot of traffic to my blog post was driven by the link that Rich Roll put on his Facebook page as shown in Figure 3. In the post, he wrote one word: Blush. Since it was a Saturday, I saw the RRR developing in real-time and it really surprised me. I happened to be shopping at a Costco Store when I got a direct Twitter message from Rich Roll himself that thanked me for the article. As soon as that happened, boom, the hammer was dropped. There were more Facebook comments on this single blog post than all of the other blog post comments I have ever received on all of my 80 something WordPress blog posts combined.

I cannot comment on the use of a Facebook page by Tableau Software because I didn’t see what happened in real-time. If Tableau is not re-posting interesting community blog posts to their Facebook page, they need to start because as you will see, there is a big difference in the RRRs of these two blog posts. I suspect that a lot of the difference in response rates might be due to the usage of Facebook. 

Figure 4 compares the first ten days of readership response rates from each article. As shown, the Tableau focused article on the Data Zone generated only about 1/2 the readership (658 vs 1237 views) compared to the Rich Roll Article. Why might this be the case? Well, although Tableau has 7,000 more Twitter followers than Rich, Rich’s followers appear to be more socially active than Tableau’s.

Figure 4 – Daily Views and Cumulative totals for Tableau vs Rich Roll Article

A simple calculation is all that is needed to investigate the difference between how active the Social networks are for Rich and Tableau. This exercise leads to some interesting insights. For Rich, the number of reads per 1000 Twitter followers is (1237/31.9) = 38.7. This means that about 39 people out of 1000 followers responded to this article. For Tableau, the number of reads per 1000 Twitter followers is (658/38.3) = 17.2. This means that about 22 (39-17) more people out of 1000 followers responded to Rich compared to Tableau.

These metrics teach us one thing: Even for blog posts that are “hot” and get re-tweeted to a lot of people, the readership rates are really very low and there is a lot of room for improvement with respect to gaining readers!  This indicates to me that using Twitter as the sole method of broadcasting my blog posts has not been a terrific choice. This is especially true since throughout most of the time I have been blogging, very few people have followed me on Twitter.  If I want to have break-out success as a blogger (which is not necessarily my intent, by the way), I will have to use another strategy than what I have been doing over the past 15 months.

 Example 2: The Growth Rate of My Blog

Since I intentionally minimized my blog broadcasting methods, what can be learned about the question I originally posed? The question is: “If you write a blog, will people find it, read it, and will your readership grow?”.

The obvious answer is yes, your readership will grow, almost by default. The reason for this is that more and more people are using Tableau software and there are a lot of internet searches being done daily to find Tableau topics of interest. If you write about topics that people struggle with, the likelihood of your blog post growing is pretty good. The primary questions I want to answer are these:

1. How fast does the blog grow?
2. Is there a limit to your daily readership response rate, or RRR?

Figure 5 shows a predictive model for my blog growth based on the number of views over time. A second-order polynomial model is used to fit the data and then to predict the cumulative number of views in the future.  If you want to learn more about how to use this type of model in Tableau, click here. A motivated reader of this post could use this predictive model to find the estimated date when my blog will hit 50,000 views.

Figure 6 shows the time series data of number of posts written, number of views, number of visitors and the number of views per visitor. Although this graph looks like an attempt at drawing valleys and ridges of East Tennessee, it does tell me some important things regarding the performance of my blog.

It looks like I have hit a monthly plateau in terms of visitors and views. The past four months have not shown a significant uptick in either measure, although Sept 2014 is going to blow all proceeding months out of the water because of this terrific post as well as the piece on Rich Roll!

Figure 7 shows that these two metrics are highly correlated, so to increase your number of views, you have to increase your visitors. If I have hit a plateau at around 2500 visitors per month, then I’ll have to do something different to increase the readership of the blog. I think this means that I’m either going to have to:

1. expand the topics that I write about to capture some new audiences;
2. concentrate on capturing more of the available Tableau audience;
3. offer free prizes to anyone willing to read my ramblings;

There is another way to increase the number of views from your blog. The obvious way is to write more articles. That doesn’t guarantee you success, however, because as I stated earlier in this piece, many people reach my blog through internet searches. They are only interested in solving the problem of the moment. They read your piece and then get back to work. It is really hard to get people to browse your blog. I know this because I have tried, very intensively to make that happen. I’ll give some examples of this in a minute. First, however, is the final graphic of this post!

Figure 8 shows the views per visitor from my blog over time. You can see that even with more content being added over time, this measure had trended downward. The more complexity that is added to a blog and the more articles that there are to read does not ensure that people will browse your content. This is a very disheartening finding and one that is going to take some effort for me to change. When I find the solution to this, it will be a great day.

I said that I have tried very hard to get people to explore the content on my blog once they arrive. Blog exploration isn’t happening very often because of what Figure 8 shows me as well as what I have learned in this experiment.

I know that intra-blog exploration is not occurring because of a few strategies I have employed during this experiment. First, I have structured the blog to resemble a website, with menus and easy navigation through word clouds and other clickable links. I have even tried rearranging these elements to gauge their effectiveness. I have put an exhaustive number of hyperlink references between the posts I have written to encourage additional exploration of the topics. I have given people a way to directly subscribe to my blog to lure them back for more. I have given away secret techniques for staying in touch with Tableau-focused blogs. I have written about a wide variety of topics to capture a wide audience. I have done deep into some topics and I have stayed shallow in others and resolved some common issues. I have also built and deployed a number of Tableau Public workbooks to track and measure user interactions. These dashboards include the following:

1. The Tableau knowledge base dashboard;
2. The Tableau Twitter dashboard;
3. The Search my Blog dashboard;
4. The Best of the Tableau Web dashboard;
5. and dozens of others that are linked to the various articles I have written. 

The quantitative assessment of these strategies are beyond the scope of this post but will likely show up in a future post. I probably will write another post specifically on the failure of this technique to keep readers interested in exploring a blog once they arrive. Once I solve this problem, Tableau will definitely be calling me to tell me that I have to work for them.

Finally, although long, this quantitative assessment is only a fraction of what I have planned to write about. There is a lot more for me to say but for now, I’ll save that for part 3.  Thanks for reading and now have a listen to the late – great Dan Folgelberg singing about Lessons Learned (just click the pic).

 

This is just an extra from a great singer.