In some of my recent posts, I’ve talked (and visualized) about some of the reasons why I love the Tableau community. My first post about the topic was a “thank you” to the Tableau community in general while my second post analyzed community member stats. Continuing with our theme, I wanted to look closer at one of the most valuable aspects of the community: the ability to find useful information. Communities like these were designed to be places for users to share information, and they predominantly come in the form of question and answer threads. The Tableau Community is no exception. In fact, few communities do Q&A better.
Case in point, if you’re doing a search on a sticky problem in Tableau, the SERPs almost always bring up posts from the Tableau Community (make sense, right?). But what’s far more impressive is that when such a search doesn’t turn up the desired results, you can take your question to the Tableau forums. I frequently tell people that if you have a question and post it there, you’ll likely get a very quick reply. So, I decided to take a look at the data behind the forums to see if my recommendation is accurate.
Quantifying the Recommendation
Up to now, I have made my suggestions based on my personal experience of receiving good answers very quickly. As we all know, personal experiences can vary greatly, and determining whether an assumption is accurate most certainly shouldn’t be based on one or even a few personal experiences. Simply put, as a data guy, anecdotal evidence isn’t enough. I want cold, hard facts! That means we need some quantification to support such a qualification. To achieve this, I wrote a Python script to pull thousands of posts from the Jive API going back almost a year. Here’s what I learned from that data:
Before gathering the data, I’d tell people they could very likely expect a response the same day. But did that end up being true? As you can see in the viz, not only is that true for 90% of posts, it’s also 77% likely they’ll get a response in an hour or less! That is fast. I’m glad to see this data backing up my personal experience of an active community that responds very quickly to questions. Working in data, you always expect there to be some sort of adjustment between expectation and reality when checking your hypotheses. In this, case the reality exceeded my expectation, which makes me that much prouder of the Tableau Community.
More Insights about the Data
When analyzing something like the Tableau Community, there are some considerations to make when choosing what metrics to lock in on and how to present said metrics through a visualization. Here are some quick insights in to my thought process when deciding these things:
- I chose to measure “replies” instead of “answers” because marking a post as “answered” requires additional action on the part of questioners. This makes it very likely there are many posts that are “answered” but aren’t marked as such.
- The top 10 tags are actually groups of tags. For instance, “filters” includes the tags: filter, filtering and filters.
- The spike in activity at 5 p.m. Pacific Time corresponds to midnight GMT. I suspect that’s due to a default timestamp assigned to some posts, not because of actual increased activity at that time.
Notice any other patterns here? I’d love to hear about what stands out most to you and any other impressions. New perspectives are always welcome!