In my mind, the result I’m about to show you from my Tableau/Alteryx blogging experiment is very interesting. I’m still thinking about what it means and how I can improve as a blogger based on this insight. I have only one plausible explanation for this result, and I’ll give you my explanation at the end of the article.
I’ve been conducting my blog experiment for about 2.5 years. I have employed pure randomness into the experiment. I publish whatever I want, whenever I want. Nobody has influenced what I did in terms of content or direction. There has been no favoritism in picking the topics of the blog posts. They just emerge in my brain and I write about them. Since my blog wanders across multiple topics over time, what I have published has definitely been purely random.
One of My Original Goals In Writing This Blog
One of my original goals was to get to about 150 weekday articles to see if it mattered what day of the week I published the articles. This took a lot more time and effort than I expected it would. With an average of about 30 articles per weekday, I thought that my sample size would be large enough to determine the result. Maybe it is, maybe it isn’t. I’ll let the statisticians decide.
Based on my previous analysis, I determined that there should not be any difference in the article performance metrics based on which day an article was published. There are at least three solid reasons for this:
- About 75%-80% of the daily traffic to my blog comes through Internet searches, so most of the articles people have read have been due to topics being searched (which is not related to the day of the week I wrote something);
- My topics were randomly generated and randomly published;
- Sufficient time has passed to wipe-out any biases that I might have developed for publishing articles on a particular day.
My expectation is that since what I have done in publishing this blog is random, I should see randomness in the results of my experiment. So let’s see what we find when I look at some article metrics which measure how often the articles have been read since they were written.
Figure 1 is a plot of the number of articles published by day of the week. It is clear that I published about the same number of articles on any given weekday. I will infrequently publish things on a weekend (about once every 2 or 3 months). Remember, this took 2.5 years to accomplish and I wasn’t keeping track of the number of articles per day in any way, so the weekday publishing results are pretty consistent with randomness.
Figure 2 is a plot of the number of article hits per day and Figure 3 is a plot of the average number of hits per day. These data represent the number of times that articles have been read over time. So for example in Figure 2, all articles written on Tuesday have been read 34,122 times since they were published. This doesn’t mean that they were read 34,122 times on a Tuesday – it means they were read that many times on all days. In fact, most articles hardly get read on the day that they are published. It takes time for any article to gain traction with the reading audience.
Therefore, this analysis isn’t about when people have read the articles, it is about when the articles were originally published. This distinction is important to understand to comprehend the significance of what I have learned.
The results shown in Figures 2 and 3 are not consistent with what I expected. I expected to see equal number of hits per day and total hits, irrespective of the day of the week the article was published.
I have to wonder how this could have happened? How could the Tuesday articles be significantly outperforming the other articles published on the four other weekdays?
Figure 4 shows the Top 15 articles ever published in this blog. It is obvious that the top 6 articles, as well as 7 of the top 10 articles, were published on a Tuesday.
Although I’m not a statistician, I don’t think that is random. I’m pretty sure that random theory would suggest that ignoring weekends, each of the five weekdays would have 2 articles in the top 10. The fact that 7 Tuesday articles show up in the top 10 is very interesting to me. In fact, if I remove the newest 6-day old article (Plot Every Nth data point), 7 of the top 9 articles were published on Tuesday.
My Hypothesis As To Why Articles Published On Tuesdays Are My Best
The answer is obvious! I simply write better on Mondays and publish these on Tuesdays! Well, that doesn’t make a lot of sense because I am sure I write the same way regardless of what day of the week it is. So now I have to move on to an alternate hypothesis.
Maybe I create good ideas over the weekend, critically think about them and begin writing about them, and then review them on Mondays. After the Monday review is complete, I publish them on Tuesdays. Although I have no data to indicate that this is what happened, I think this is likely the reason why this non-random behavior has occurred. I believe I can elaborate even further on this hypothesis.
Over this time period, I have been deeply immersed in analytics. My brain has been firing seven days a week as I think about ideas and then try to implement them. All you have to do to confirm this is to ask my wife about how distant I can be when I’m immersed in solving a problem.
I have long understood that my brain can work on unresolved problems in the background, while I’m attending to other tasks. The first time I recognized this is when I solved the equation for the mass of the universe – while I slept. I awoke the next day, wrote the equation, solved the equation, and then checked the answer (I was right!). This was the first time I ever remember doing such a thing.
The next time I did it was even more compelling. It was in the late 1980’s or early 1990’s and I had spent several months writing the computer code for a contouring algorithm I created. The code I was creating was designed to be a fast and efficient method for for contouring results from numerical models. Back then, optimized code was needed because we didn’t have the luxury of cheap memory and an abundance of computational speed.
As the code complexity grew, so did the book-keeping needed for the method. Although the method was simple in design, my implementation of it grew too complex. So one morning after months of writing code, I was in the shower and the complete, simplified solution appeared in my head. I can still see the moment in my mind, as though I’m watching a commercial on TV from that time. I went to work, scrapped months of code, and wrote the majority of the new solution before lunch.
So what does this have to do with my best articles being published on a Tuesday? It is simple, really.
Good articles take time to create. The ideas have to good. Your insights to the problem have too be sound. The explanations have to be good. People have got to want to read about the topic to generate sustained interest in these articles.
Mostly, the topics that form a good blog article have to be either fundamental techniques needed by everyone or they have to propose solutions to common problems experienced by a lot of people. Otherwise, people forget about blog articles and they fade to oblivion.
For an article to stand the test of time (i.e., continue to be read long after publication), the topics have got to be interesting and they have to explain fundamentals or solutions to ubiquitous problems, such that people are compelled to search for these explanations.
Therefore, my best articles have been on topics that are either fundamental techniques I explain or problems that I discovered by using Tableau on a daily basis for about 8 years and Alteryx for the past 1.5 years. I knew that the techniques were something that I had to do over and over, on job after job. The problems I resolved were also commonly occurring in most jobs.
To resolve those issues, I thought about them deeply before deciding to write about them. I took my time. Once I was dailed-in and understood the issue or technique very well, I spent the weekend writing and explaining the topics. I used Monday for a review period. I waited until Tuesday to publish the articles.
The lesson for me is that I can write good articles if I’m patient enough to take the time to describe the technique/problem, develop the solution, and then write about the lesson learned. That is what it takes, and that is why it happens for me mostly on Tuesdays.
It turns out that this process isn’t so random after all, although I was sure it was going to be. Thanks to Tableau, I could see the truth to this behavior in a couple of minutes, even though it took me over a couple of years to create the data.
Who cares about this article, anyway? Well, I do. I wrote it for myself. That is all that matters.
I’m doing this experiment to learn how I can help the most people in the shortest amount of time. If I can optimize my blogging creations, I can help more people. I’d much rather write one great article rather than eight so-so articles. If I didn’t take the time to look at this data, I never would have found this insight. That is one of the reasons why I’m doing the experiment – to learn how to be an effective blogger!
In this case, I happened to teach myself something that might also help other bloggers if they care to pay attention. This is one lesson that will help me select and refine what I write and hopefully to eliminate the insignificant noise. My blog should get better because of this.
In an upcoming article, I’m going to show another interesting result from this experiment. Until then, thanks for reading.