Friday, November 30, 2018

Tableau's Lost The Helping-People Path

This post is a continuation of a discussion about Tableau's seeming reluctance to improve the product's ease of use.


It might be obvious, but I'm also disappointed that Tableau has effectively abandoned ease of analysis as a primary product principle. My list is only a mere scuff on the onion skin of things that would make Tableau a much better product. Some of them are so simple and easily fixed that their persistence is irritating—it's almost as if Tableau doesn't take the things that bug mere humans seriously. I know this isn't totally true; there are people at Tableau who really do care about these things, which makes the situation more mysterious and, oddly, more vexing.

At its core, the things that Tableau makes really easy to use for analyzing one's data are the handful of operational functions—field selection, sorting, aggregation, calculating new fields, and filtering—that 4GLs made easy in the 1970s, albeit in a mainframe character-mode terminal and line printer world.

Tableau extends this model by providing visual forms appropriate to the analytical context and makes it easy to apply visual characteristics, e.g. colors; although it's always been poor at clarifying its operating principles, e.g. why does the order of double-clicking fields produce different effects.

That pretty much covers Tableau's "best-ever" space.

There are a number of dead simple things that could make Tableau better even in this space, e.g.

  • Getting rid of the default cells' 'Abc' label – it's pretty silly, since the by far dominant purpose of the cell is to contain quantities, and the 'Abc' implies that text is the normal content. Even '123' would be preferable, although it's not without problems.
  • Adding scrolling to the dimensions' part of the viz – the 'Ideas' topic that prompted this discussion.

There's a vast space for improvements in expanding the functional easy-to-do analytical operations horizon: making more things simple, easy, and straightforward.

I spent a lot of time trying to get Tableau to hire me to help, and gave up when they were pointedly not interested.

It seems pretty clear that Tableau's been working very hard to make itself into an enterprise platform - there's a Motley Fool interview with Christian Chabot from a number of years ago (can't quickly find it) where he talks about Tableau's objective being to maximize its market potential in the context of enterprise platform software sales. This has never been a secret. Even as a new company Tableau used an enterprise sales approach: when I bought my Tableau license (money well spent) in 2006 I had to purchase it from an actual salesman who was primarily interested in how many other copies he could sell into the organization I was part of.

Bald-faced self-promotion here.

A couple of years ago I took the opportunity to put together a product proposal and some initial prototypes for a visual data analysis tool targeted at individuals with data they need to understand.

It addresses Tableau's shortcomings, with the advantage of over thirty years' experience in helping people with their data analysis needs—including stints as Product Manager for PC/FOCUS and FOCUS for Unix where we created the first generation of visual data analysis tools (alas, IBI did what Tableau's doing and looked to be more 'technology' than 'human' oriented)

It took the better part of a year and I'm pretty pleased with the product concept and design. Sadly, I'm not part of the 'business' side of this business, and the limited pathways to the resources required to build a real, live tool I've had access to haven't proved fruitful.

So... I'm open to any opportunities to collaborate in creating the next great human-oriented data analysis tool. One that's affordable and makes it simple, easy, and straightforward for non-technical people to connect with and understand the data that matters to them.

Given that Tableau's history strongly suggests that it's not going to revisit its basics and improve its fundamental functionality, I'm waiting for the next great tool to arrive, one that addresses Tableau's friction points and expands the range of what's is simple and easy to do data-analytically.

In the meantime Tableau remains the best tool I've ever found for basic data sense-making and I'll keep happily using it.

Thursday, November 29, 2018

A list of ways in which Tableau can -should- be improved.

Tableau's in need of a reworking, a redesign of it's fundamental data analysis interactive model. The original design concept worked well in the original functional space, but it's been left languishing and is in real danger of becoming left behind by new innovators.

It's been quite a while since I've spend much time and effort into advocating for improvements to Tableau's support for ordinary humans' data-analytical abilities.

At this time I'd like to present a compilation of material I've previously published, augmented with a wee bit of explanatory information.

I'm motivated to do this because I hear whispers that Tableau is maybe taking a look at how they can improve things.

Here's a list of references to material describing some ways in which Tableau could be improved; it covers a pretty broad span, partly because everything's related, partly because it's worth taking advantage of every opportunity to advocate for making Tableau a better product for helping people see and understand their data.

Nuggets and Seeds
A compendium of thoughts and musings about data analysis, effective data-analytical software tools, and Tableau's position as a premier tool for helping people see and understand data.
Is Tableau in danger of becoming just another enterprise platform? I hope not and fear so.
January, 2016

Rethinking The Analysis Frame #1 - Row Folding
Directly related to improving Tableau's table(ish) abilities.
September, 2013

Precision Inputs Required In Addition To Analog Controls
Precise control of the geometries of the various structural elements, e.g. column width would go a long way to making Tableau better.
Related: make it possible to select and adjust multiple elements at once - the current select/fiddle/repeat interaction model is tedious, tiresome, boring, and error-prone.
November, 2013

Inconsistent Chart/Table Formatting
October, 2012

Towards Better Formatting - Notes on Alignment
Specifically about text alignment, it lays out a robust set of options for formatting and aligning text.
Related: text should be formattable wherever it appears, this includes field labels, where it would be incredibly valuable to have control over folding, new lines, etc.

Problematic Table Formatting When Deployed in Dashboard
July, 2013

Jittery Charts - Why They Dance and How to Stop Them
January, 2013

Failure to Identify -or- Who is that mystery measure?
May, 2013

From Chart White Space to (the need for) Architecture
General thoughts on the need to implement a coherent visual architecture underpinning Tableau's visualization space and data-analytical interactive functionality.
July, 2013

Enhanced Chart Design - Adding White Space to Bar Charts
Visually separating elements grouped by Dimension members makes it much easier to identify these groupings than Tableau's current identically-sized matrix layout paradigm.
July, 2013

Dual Axis Visibility Configuration Explained - Is Not What It Should Be
August, 2013 Dual axis charts' axes should be independently and rationally controllable. They're not.
While we're at it: axes should be top-level objects, not subordinate to Headers, which means that one needs to have the Header visible in order to manipulate the axis. Which is bad, but worse: one needs to know that's how to get to the axis, and that's an impediment for non experts.

Is it Transparency? Is it Opacity? Labeled one, works like the other.
A little thing, but important in that 1) it works contradictory to expectations (that it established), and 2) plants a seed of mistrust in Tableau (if it's wrong here, where else is it?)
December, 2013

Additional Improvement Opportunities

Needed: More File Names and Better Interaction
April, 2013

Tableau Needs a New Windowing Scheme
(a bookmark article)
Essentially, Tableau's current application UI is horribly constrained by the fixed configuration of its component windows/panels.
The UI architecture made a little sense when Tableau was initially introduced, a legacy of the project Poseidon UI, but it's been outmoded for a decade.
Modern application UIs are much more modular and flexible, letting one have the tools necessary for the job at hand readily available without opening, closing. collapsing, expanding structural, modal components. March, 2014
Speaking of modality: the mishmash of modal/non-modal sub-windows and dialogs is a mess. Wherever possible, dialogs need to be non-modal.