One Year Later – New Predictive Analytics for My #Tableau and #Alteryx Blog


It was about 2.5 years ago, when I started writing this blog about Tableau and more recently, about Alteryx. One of the reasons I wanted to do this was to conduct an experiment on the growth rate of a technical blog. I wanted to answer this question: “If you build it, will they come?”.


I have previously written articles about the readership growth rate of my blog. Here are a few references to this material if you are interested:

  1. In a two-part series, I wrote about Lessons Learned as a Tableau Blogger (click for Part 1 or Part 2),
  2. An update to that work occurred late last year where I specifically developed a predictive model for my blog readership growth (click here). In this article, I wrote the following statement:

I think that it is going to be really hard to maintain a second-order polynomial growth rate for this blog because I only publish original content and do not depend upon the work of others to build readership. So for the record, Figure 4 shows the prediction for blog growth over the next 12 months. That trend is going to take a lot of work to achieve!

With another year passed since I first established that model,  I wanted to check-in to see what has happened since I wrote those words.

Original Polynomial Model Results

Figure 1 shows that the original polynomial model was accurate for about 9 months after it was created. By June, 2015, the blog growth rate slowed such that the original model was no longer giving accurate predictions of blog readership.

Predicted Vs Actual Readership

Figure 1 – The original polynomial model and the actual blog readership numbers.

To understand why the polynomial model lost accuracy, I took a look at my posting frequency. Figure 2 shows the number of articles I published over time for the past year. Due to starting a new job in February, moving, traveling and other activities, the number of articles I wrote was fewer than I had done in the past. In particular, I didn’t write much from mid-March to mid-July, which was a four month quiet period.


Figure 2 – My blog posting activity from October 2014 to the present.

For this reason, the polynomial model became inaccurate by early June, 2015. In effect, the polynomial model lost its accuracy because I changed the frequency of the content I was creating.

Model Recalibration

As I had predicted, maintaining a second-order polynomial growth rate has not been possible. By changing the Tableau trend model to a linear model, a new predictive equation that better fits the data was achieved. Figure 3 shows this new model.

Sept 9, 2015 - New Linear Model

Figure 3 – A new linear model that can be used to predict the readership of 3danim8’s blog.

Each blue dot is actual blog readership numbers over time. I try to capture the data about every time 1,000 more views occur on the blog but sometimes I miss recording the data. For the record, this model predicts a total readership of 136,813 on 12/31/15. I’ll have to check on the accuracy of that prediction on New Years Eve.

Insight on Weekdays vs Weekends

This model slope indicates an average readership of 225 articles per day. One of the interesting insights I have seen, however, is that my blog is definitely a weekday type blog. Figure 4 shows the recent daily activity.  The weekday readership totals are generally around 300 articles per day but the weekends are 100 or less. This confirms my suspicion that people use my blog to solve problems during their workday, which is exactly one of the reasons why I write the blog.


Figure 4 – Daily blog readership numbers. Note the repeating pattern of 5 strong (weekdays) days followed by 2 slow (weekend) days.

Final Thoughts

Quantitative experiments like this take a lot of time to create (writing content) and a lot of time to develop (watching what happens over years). So what is the value of such an experiment?

For me as the author of this material, I have learned that writing 3danim8’s blog really helps me perform my professional job with greater speed and accuracy. It also helps me in other ways by keeping me aware of new software technologies and by learning new techniques for solving problems. For these reasons, this experiment will continue as long as I can create new ideas to write about.

For consumers of the information, they have free access to problem-solving concepts, techniques and real-world examples to help them learn the software. The material included in this blog is intended to help readers learn new things. In particular, I focus on writing problem-solving techniques by providing complete, real-world examples to show how complexity in data is handled efficiently to create insightful visualizations.

For the software companies (Tableau and Alteryx), the monitoring of blog readership can provide insights to help them understand what customers are struggling with. If particular topics remain salient over time, it means that the software companies have not addressed deficiencies in their documentation or the structural design of their software.

For those reasons, a long-term experiment like this can provide benefits to many people in multiple ways. Thanks for reading.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s