Meta Tableau

My Biggest Tableau Desktop Headaches (and Probably Yours Too)

I love Tableau. It’s my hammer that makes everything look like a nail. No it’s not right for everything, but it can do a lot of things really well once you know how to use it. And because of this loving relationship, I also know exactly what I hate about it. Like an intimate relationship with a lover, the overall experience with Tableau is something you want to shout to the world. But also like an intimate relationship, Tableau has those small things about it that really grinds your gears. Kind of like when someone you love makes a little annoying noise after they drink a beverage. Technically it’s not wrong in any way, but you just want to take away every single beverage from them for all eternity so they can never make that noise again. Ok… maybe that was a little far. Anyways…

Tableau Desktop has been the data visualization tool of choice for me for the last 7 years. It has grown from a pretty good tool into an excellent platform. Tableau Desktop in 2020 takes care of so many of the gripes that existed a few years ago in previous versions. Things that were once nearly impossible are just a click of the button. Things like Set Controls and Relationships. Tableau has provided more tools for every step of the process, from simple data blending to completely managed server environments. That’s without even getting into extensions, APIs, etc.

I love Tableau and believe they will continue their track record of implementing user-desired features. Tableau has been one of the best companies I’ve interacted with, who truly takes user feedback to heart and continues to make their product more enjoyable to use. So while these are my current 7 biggest headaches, I can’t wait to see what the future holds for their platform.

My 7 Biggest Headaches

1. Being unable to modify the layout hierarchy from the dashboard layout tab

Tableau has a beautiful layout pane to select objects in the dashboard. You can adjust padding, margin, background color, etc. from here. But you cannot rearrange the objects using the hierarchy displayed in the layout pane.

It feels natural to be able to click and drag an object in the hierarchy and put it somewhere else. This is not the case though, you can only do it on the dashboard itself. If and when this feature comes out, it’ll cut down on dashboard building and reorganizing times and headaches significantly.

2. Not having a PROPER() function

We’ve got UPPER()… We’ve got LOWER()… We do not have PROPER() or TITLE(). Sometimes you can’t control your data source because you don’t have the right permissions. But you need to change a field that’s in all uppercase or all lowercase to something that has the first letter of each word capitalized.

This doesn’t exist in Tableau. It exists in some form in Excel, C#, Python, Java, and more languages. This fact paired with headache #3 makes title/proper cases in each workbook painful.

3. Related to number 2; not being able to define our own functions

Ever use the same calculated fields and functions over and over. Each and every workbook. You open it up and you create another calculated field that does the exact same thing in each workbook.Long calculated fields, like your own custom version of making the proper case from point number 2.

If custom user functions were possible, users wouldn’t have to ask for specific functions to be added. The flow for a user could be much more efficient too.

As a counterpoint from Tableau’s standpoint, allowing users to manage and implement their own customizations like this can do a few negative things. Things like new support requests claiming a bug because a user can’t figure out why their function doesn’t work as expected. Or that this type of functionality would drift the product from user configuration and light development to something more of a user development tool. This could potentially shift the target user for Tableau Desktop, and I doubt Tableau would want to drift from their current pinpointed target user. One last point is that allowing user defined functions could potentially open up security issues. It’s just a whole new bag of worms.

Still, the pros for usability outweigh the cons for me.

4. The limitations of organizing worksheet tabs in workbooks

Simply, we need a way to group and organize tabs differently. The colors are nice, the toggle to change the tab layout is nice, but scrolling left to right constantly for large workbooks is not nice. If you could group tabs into folders or hierarchies just like calculated fields, that’d be great.

5. Restrictions on bulk editing dashboard objects

As of 2020.2, users are able to only format the standard configurations of categories of objects in bulk. For example, you can format all parameters in the workbook by changing its font, color, borders, etc. What you can’t do is select 3 dashboard objects (like 3 worksheets) and adjust all of their padding or margins. Instead, you have to select each worksheet on the dashboard and edit its settings one-by-one.

My back of the napkin calculations estimate that bulk editing of dashboard objects would save me about 15 minutes for each dashboard I have ever created.

6. Lack of hierarchical filters

Hierarchies are a feature for building dashboards already. When you establish these hierarchies in the data pane, it allows users to expand axes to drill up and down hierarchies. A common request and question I get from end users is, “why can’t I drill down in filters?” Or something similar to that.

You can place all of the separate levels of hierarchies as their own individual filters, but cannot have a singular filter where you can select parents and children, or expand the hierarchy to drill down to other levels. From a user’s standpoint it makes sense. You would expect a relationship like a hierarchy to function as a hierarchy on the graph portion, and in the filters. Something like the below, taken from the Tibco Spotfire documentation:

7. Formatting in tooltips

There’s a workaround for this one, but it’s not incredibly customizable and has it’s own limitations. That workaround is creating embedded sheets in the tooltip so that you can do things like control the background color. But to create a custom sheet for every tooltip you wan’t to customize is exhausting and adds baggage to the workbook.

Being able to do some basic things like change tooltip background color and transparency would go a long way. Making the tooltip almost like a page builder or able to accept HTML would be spectacular. I won’t get my hopes up for that last one though, as it would be a huge undertaking for Tableau to develop and support.

Wrapping It Up

Tableau Desktop is a spectacular tool. Whipping up interactive and insightful visualizations for users is 1000x easier than what was around before Tableau, and it’s truly led to a democratization of data visualization and analysis. With that comes growing pains and millions of change/feature requests. These are 7 of mine and maybe some of yours. Selfishly I hope these are the next 7 features on Tableau’s product map.

I’ll soon be posting my favorite features of Tableau since version 2019, as well as my headaches/favorites of Tableau Server and Tableau Prep. So keep a look out for those.

P.S. I had to update the title to reflect that this is specifically about Tableau Desktop. Since Tableau has been expanding its platform which includes more advanced Tableau Server capabilities and an online visualization builder. These offer separate functionalities compared to what I’m talking about in this post with Tableau Desktop.


Data Visualization is Not Even 25% of the Work

What Did You Just Say?

I’m a Data Visualization specialist at this point in my career, and I’ll tell you an unpopular opinion… The visualization part of data projects really isn’t the hardest part of the project, it’s not the most important, and it’s the least time consuming. Even with all of these factors considered, it’s often the most visible and emphasized part of the project (no pun intended) in many businesses. This is wrong.

I’ve estimated that not even 25% of the work on data visualization projects has to do with the visualization. Are there any studies or hard numbers to back this up? Nope. Just an estimation which I’ll go through now.

Estimating a Data Project Timeline

When it comes to a full-scale data project that ends in a visualization, the hard work and complexity happens behind the scenes. Gathering business objectives, setting scope, listing deliverables, data collection, data exploration and availability, data cleaning, data structuring, exploratory analysis, and maybe some additional modeling and data science before we even begin crafting an end visualization. That right there is up to 10 different steps and could be broken down further. If we (wrongly) assumed that each step took the same amount of time, then data visualization only takes 9.1% of the time (1 out of 11 steps).

In reality, the longest and most difficult portion of the project happens in the “unsexy” part. That would be data collection through data structuring. These steps are the bulk of the work. Unless the data set you’re working with is simple or there has already been significant effort in building clean and structured data sources, you’ll spend significant amounts of time exploring and verifying data. The reality is usually about 30-70% of the project timeline is spent in these phases. The other reality is that you’ll go through these phases of the projects several times, since the first few presentations of the data visualization will bring up many more questions on data quality and how the project ended up with the values presented in the data visualization.

What is the amount of time then for the data visualization part? If you have your business questions laid out during planning and you are aware of the analyses that need be done, there are only so many paths to take. There are a limited number of visualizations you can choose. The meat of your data story is already outlined with the business requirements. Sure, you can spend more time digging for additional data stories to tell, or on the UX/UI components to make it look better. But additional analyses are cherries on top of the outlined deliverables. And once you meet a certain design threshold, a slightly better look won’t fundamentally change how the visualization is received.

Why Isn’t The Data Visualization the Most Important Part?

Ok, maybe I was a little harsh. The data visualization portion of a data project is important. Striking out on the data visualization can make a project bust. But an excellent data visualization can’t make up for a poorly executed planning and data collection phase. It can’t make up for bad data science or inaccurate data sets. That’s why it isn’t the most important part. What a good data visualization can do is: surface bad data, surface inaccurate data science methodologies, answer business questions that should have been part of the planning phase, and much more. Yes, data visualization is important in data projects. Maybe 2nd most important, but it’s not the be-all and end-all.

But That’s My Job…

Building impactful data visualizations can provide great value to your organization and data projects. So being efficient and good at what you do can certainly provide great value to your company, job security, and a fruitful career. That being said, if you want to make your job more resilient to economic forces, you need to keep some things in mind.

If you’re a Data Viz specialist like me, you need to constantly work at delivering additional value outside of simply your visualization. The one exception to this would be a consultant hired specifically for data visualizations while everything else is perfectly prepared (ha!). Even as a consultant you’re arguably always better off delivering more value than anticipated for your customer.

But I digress. Deliver extra value. Somehow, some way. Whether it’s through your data visualization or through other parts of the project you’re working on. Having a diverse skillset that provides a significant multiplier of your cost to a company will make you a prized member of your company.

For some this is easy. Maybe it’s because you’re actually a BI developer, which encompasses many more responsibilities. Maybe you’re simply the data person in a smaller organization and therefore handle more of the data pipeline. If you’re not explicitly in a situation that pushes you outside of data visualization, just beware the fragility of your position and the need to diversify your value contribution in order to make your position more resilient. Just because you’re job box has a label doesn’t mean you can’t break out of that box or just pop the top open and relabel it yourself.

Give Me A Parting Analogy

Constructing a building is a great parallel to a data project and its steps. The planning, permits, surveying, laying of the foundation, erecting of the frame, installing the mechanicals, and putting up the drywall and insulation take the bulk of the time. Especially before the frame goes up, how many construction projects have you looked at and said ‘they never make any progress on this, they’ll never finish!’ This is only before you return a couple weeks or month later and everything is finished! It’s not that nothing was happening before that last stretch of time, it’s just that the progress wasn’t visible to the untrained eye.

This is equivalent to the business planning through data structuring part of a data project, especially from an end-user’s perspective. Nothing happens… nothing happens… then boom! The data visualization presents itself with all the work wrapped up into the visible end-product. Like the finishing steps of a construction project, the data visualization is what everyone will see and pay attention to. As long as there aren’t horribly obvious mistakes or incredibly artistic/unique details, most people won’t be too moved. The end-product will simply serve its purpose.

On the flip-side, if the foundation and frame of the construction was poorly or improperly done, the looks and/or functionality of the interior and exterior finishes won’t matter. In fact, the construction will eventually become unusable due to its improper foundations. Data visualizations are no different. Without high quality data, structure, and planning, the whole thing falls apart. Then your visualization answers the wrong business questions or answers the right questions incorrectly.

So remember, get the foundation right and spend the most time on it. It’s the most important part. Then focus on the visuals, because that’s what the people will appreciate.

I’m always looking for feedback, tell me what you think of this post! – Dan


Success! You're on the list.