Additional Insight and Clarification of #Tableau Exponential Trend Models

Introduction

Earlier this week I wrote a post about how to understand all the Tableau Trend Models.  This is a follow-up to that post.

There was one lingering issue that I didn’t resolve before publishing the information (life isn’t perfect!). This issue was the disagreement between the Tableau published intercept and the correct value of the intercept for the exponential trend model.  After some additional thinking on the topic, I was able to resolve this difference.

Debugging computer code is something that comes naturally to me.  It is a little more challenging (OK, maybe it is even impossible) to do when you don’t have the code in front of you to examine.  However, sometimes you get lucky, or sometimes you just know that there is a way to figure it out.  I recently wrote about an interesting example of this no-code debugging capability.  I think you get good at things like this if you write a lot of computer code.  In any event, I realized that if I did a little more testing of the Tableau exponential trend models, the truth would be revealed.


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!

Enter your email address to follow this blog and receive notifications of new posts by email.


Methodology

I used the same method as before, but now I created eight exponential models, each having different intercepts and growth rates.  Four of the models were in base e and four were in base 10.  Figure 1 shows the curves for the four base e models.

4_exp_models_base_e

Figure 1 – Four exponential models used in Tableau (click for full size).

Analysis and Conclusions

Table 1 contains the results of the eight tests.  I didn’t resolve the intercept discrepancy until I looked at test case 3.  In this test case, a negative intercept was used to mirror (i.e. flip the curve about the y-axis) the example used in case 2.  When I saw that Tableau could not complete a trend model because the y-axis data was negative, I realized the solution to the problem.

Exponential_Math_Models_Results

Table 1 – Results of 8 exponential trend models (click for full size).

This behavior indicated to me is that Tableau must be using a log-based, linearization methodology for the exponential trend models.  Since Tableau could not produce a trend model for test case 3, as shown in Figure 2,  I realized that another type of log conversion would be required to get the correct intercept.

Test_Case_3_Base_10

Figure 2 – Tableau was unable to complete a trend model for Case 3.

What this means is that Tableau has simply forgot to tell us to take the anti-log of the intercept to get the correct value, as shown in Table 1.   Therefore, if you use exponential models, you have to pay attention to a couple of different log based conversions that will be needed whether you are in base e or base 10.  If you are in base e, you need to take an anti-log of just the intercept.  If you are working in base 10, you need to do a log scale conversion on the slope followed by an anti-log conversion (base e!) on the intercept.  I hate it when you have to mix and match bases like that!

To end this trend model confusion, Tableau need to do the following four things at a minimum:

  1. They need to fix their trend model code so that we don’t have to do log-based conversions.  They could be even friendlier to us by adding a button for base e, one for base 10, and one for base anything you want it to be!
  2. They need to improve the documentation on the trend models by improving the naming conventions (e.g. call a slope a slope, not an x-value), and by providing some examples like I created in these blogs,
  3. They needs to provide the correct governing equations for their trend models (all of them are technically incorrect at this time, some more so than others!),
  4. Allow us to gain access to the trend model terms such as r-squared, slope, etc, so that we can post these variables onto our plots!

I love Tableau, but sometimes you have “gotta tell it like it is” to make things better for all of us! At least now I know how to confidently use all the Tableau trend models, which has been something that has bothered me for seven years. Many years ago, I contacted Joe Mako about a problem related to these trend models. He had to punt at the time, just like I did, because the answers weren’t abundantly clear and neither one of us had unlimited time to investigate my question. That should tell us something, because as we know, Joe is one of the most highly regarded Tableau Masters! End of story and I promise to leave these models alone!

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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