1 2 Previous Next 17 Replies Latest reply on Oct 2, 2015 7:54 AM by Simon Runc

    How Do You See Tableau's Canvas?

    Rody Zakovich

      There are two things I get really nerdy about.....Physics Documentaries and Tableau. Interesting enough, these two worlds recently collided.


      I was watching a Documentary on Parallel Universes and I began to think, "Hmmm, what would happen if I thought about Tableau like this?". Where Tableau is not a 2D space, but a 3D space comprised of multiple layers stacked on top of one another. Each layer takes on the properties of the previous (All of which are bound to Tableau's universal law [Order of Operations]).


      Now I don't know if this idea is valid (I still have a lot to learn), but it made me start to question how I view Tableau, and I think that is most important.

       

      So, now that I have gotten that out of the way.....I wanted to ask others how they see Tableau.

       

      There is a great TDT that discusses this topic, as well as links to amazing visualizations (All that required "Out of the box" thinking).

       

      TDT: I didn't know that was Tablossible - August 27, 2015

       

      How do you see Tableau?

       

      Do you think of Tableau as a 2D model, or something more abstract?

       

      How does this way of thinking affect how you approach complex problems?

       

      As always, any feedback/thoughts are highly appreciated!

       

      Jonathan Drummey Simon Runc  Noah Salvaterra Matt Lutton Bill Lyons Bora Beran Mark Fraser Jeff Strauss

       

      Best regards,

      Rody

        • 1. Re: How Do You See Tableau's Canvas?
          Matt Lutton

          I see Tableau as a drawing tool, primarily, without many limitations - with awareness, we can produce nearly any visual result we want to represent our data, given the correct data set(s).

          2 of 2 people found this helpful
          • 2. Re: How Do You See Tableau's Canvas?
            Simon Runc

            Hi Rody,

            ….Interesting question, and approach! I too am a bit of a wannabe ‘Theoretic Physicist’…As there are a few different ideas on Parallel Universes, l’m going to list out the ideas I’ve come across.

             

            CAVEAT ALERT: I’m a total hobbyist when it comes to Theoretic Physics, so the below understanding is from that perspective…so would be delighted to be corrected if any of my understanding is incorrect!! – which is pretty much a given!

             

            Parallel Universe 1 – Part of String Theory predicts there are actually 9 dimensions, with the 5 we don’t experience are ‘tiny’ curled up dimensions, where string interactions within these ‘rolled-up’ dimensions affect our ‘experience-able’ dimensions

            Parallel Universe 2 – Space (as we know it) is infinite, as there are a finite number of ‘fundamental particles’ there is a finite number of ways to ‘arrange’ these particles….if you travel far enough (10^118 meters is a number I’ve heard…which btw if you tried to write out and put a zero on every atom in the visible universe, you’d run out of atoms!) you’d run into an exact copy of yourself!

            Parallel Universe 3 – Similar to 2, except that this time multiple universes ‘inflate’ (like bubbles) in a Multi-verse medium (sometimes called a ‘Brane’), similar to ‘Swiss Cheese’. As this, ‘medium’, is infinite, so are the number of ‘alternate universes’.

            Parallel Universe 4 – Everytime a decision is made, quantum theory predict the universe splits and both decisions are realised in their ‘own version’ of reality. These universes exist parallel, but without interaction, with one another.

            Btw there is no reason, why all 4 couldn’t be simultaneously the case!...Love this stuff although it does make my brain hurt…in a good way!! btw if anyone else is interested in this (and loves a bit of 'brain hurt'!! 'Our Mathematical Universe: My Quest for the Ultimate Nature of Reality' by Max Tegmark is a great read.

             

            The reason I detail these out (apart from the fact I’ve never had the opportunity to write about Parallel universes before, which is the main reason!), is there is a difference between parallel universes which interact with one another, and those which exist in isolation of one another…I’ll come back to this later!

             

            So in simple terms (and this is something I really try to stress during any training…and is very well expressed in the ‘old-but-great’, Blue-Things-and-Green-Thing http://www.theinformationlab.co.uk/2011/09/23/blue-things-and-green-things/)...

             

            I see Tableau as a ‘2D Drawing Canvas’. Whenever I add a Blue (discrete) Pill onto the Column or Row Shelf this will split, equally, the canvas by the number of elements in the Pill (eg. I have a Blank worksheet canvas, I drag a discrete-dimension with 3 elements onto my Column Shelf, and I split my canvas into 3, vertically). Within each of these ‘Canvas Segments’ (or total canvas if I have no Blue-Pills), for every Green-Pill I add I get an Axis, allowing me to ‘plot’ a mark for the measure at that location. The number of marks (and thus how they are calculated/aggregated) is determined by the ‘VizDetail’. I really try to stress this ‘understanding’ is better than ‘learning’ approach as it’s more interesting (for one), and means you should be able to work out how to display, almost, any visualisation you desire (as the Tablossible TDT shows so well). This is looking at it as a single canvas…this could all be summarised by the ‘Blue-Green-Things’ article ‘mantra’ that, 'blue things create headers and green things axis'

             

            I extend this ‘model’ further when I use ‘Dual Axis’, and/or ‘Reference Lines/Bands’ to be ‘multi-layered’. Dual Axis and Reference Lines give me 2+ (I put ‘+’ here as I can have many Reference Lines/Bands) further layers. Is see these layers as separate (as in they don’t interact with one-another) and akin to GIS/Photoshop layers, where they are independent, but are overlayed. This ‘analogous’ to either of the last 3 parallel universe models (this isn’t a perfect analogy as ‘VizDetail’ can affect Reference lines).

             

            The next part, which my current thinking is closer to Parallel Universe Theory 1, is how I think about calculations…I’m also far more hazy here!...and really looking forward to hear some others ideas on how they ‘conceptualise’ this. I could see a useful ‘conceptual’ model based on this idea, being the various calculation type (Row Level, Aggregated, Table Calculations, LoD) as layers, with the Order of Operations defining the ‘Universal Laws’

             

            …as I say still thinking about this, so very interested to hear others, ideas on how they ‘conceptualise’ this in an ‘abstract’ way.

            2 of 2 people found this helpful
            • 3. Re: How Do You See Tableau's Canvas?
              Rody Zakovich

              Thank you Matt and Simon for contributing your thoughts.

               

              Simon, I very much like your approach. The article from The Information Lab was one of the first Articles I read on the Tableau Canvas, and it is still very valuable.

               

              Like you said, calculations are really where I start to see this "layering" in Tableau. I try to imagine each calc, as a new layer. The exception to this is FIXED LoDs and Standard Row Level Calcs (Which I imagine being in the 2nd layer [The first being the Datasource(s)]).

              • 4. Re: How Do You See Tableau's Canvas?
                Bill Lyons

                I guess my answer is, "it depends," so I guess, "generally," Relativity is involved.

                 

                When I first got Tableau, it was a "Big Bang" experience for me: I was in a whole new universe. The first couple of months, it was a 2D universe, but I rarely think of it as 2D anymore, though for some simple projects, I do. With dashboard actions, etc., it becomes 3D. The addition of time in page animation produces 4D (I will be SO HAPPY if they ever figure out how to do that in Server!).

                 

                But all of that was just the beginning. Then, when I began exploring Table Calcs, I felt like I was on the event horizon of a black hole. Now, Table Calcs and LOD expressions sometimes seem to produce effects akin to quantum entanglement and "action at a distance," which is awesome! Still, when I see things like the TDT Tablossible sessions, and the posts of Jonathan Drummey, the Uncertainty Principle is reconfirmed, and there are still times when I feel I am "in search of Schrodinger's cat."

                2 of 2 people found this helpful
                • 5. Re: How Do You See Tableau's Canvas?
                  Simon Runc

                  Great stuff Bill...Some nice 'Quantum Name Dropping' there!! I'd also add the 'Observer Effect' to the list with regards Table Calculations...in that we don't know what value a Table Calc will take until we measure it (aka Add it to the Canvas!)...I like this analogy (and explains why Table Calc don't appear in the Data View screen). I get quite a few questions from people on why, once they have table calc working as they want, and try to use it in a different worksheet (with different dims/grain) it no longer does what they want...I might try this analogy on them next time!!

                   

                  Rody,

                   

                  I have been thinking about this (let me re-phase that!)...I have been 'over-thinking' this the last couple of days! as I think this has legs...My current attempt is to try and 'follow' a datum through it's journey from Row Level, to the other states (universes) it can inhabit, what restrictions to travel it has, and how its identity is affected along the way.

                   

                  The below image is my 'work in progress'!! So if we take our Mr. Row Level (a Datum, who is native (and a national) of 'Row Level World'). He can move to Aggregate World, where he would be 'nationalised' with a new Aggregate Identity, he can also embark down the 'FIXED-LOD' path, but this paths returns him (aggregated with some of his friends!) back to Row Level world, although again he needs a new identity (he's not the same Datum that left all those seconds ago!!). From Aggregate World, one can then move on to 'Table Calc World'. Notice that there is no path from 'Row Level World' direct to 'Table Calc World', and this analogy extends to why you can't mix Aggregates and Non-Aggregates (these are not very progressive worlds, with no multi-culturalism!!). What I need to do next is somehow, use the 'Order of Operations' on the right, to show how various 'Operations', and their order, (along each path) affect the journey. So for example, with the FIXED-LoD path, the datum would travel un-affected through Dimension Filters, but would if passing through a Context Filter....btw I've not even attempted to think about INCLUDE/EXCLUDE yet!!

                   

                  As I say, still very much a 'Work in Progress'!!, but interested to get your thoughts...right better get on and do some actual work

                  Tableau Calculation Medium.png

                  2 of 2 people found this helpful
                  • 6. Re: How Do You See Tableau's Canvas?
                    Bill Lyons

                    Looks like you are on your way toward a Unified Theory of Tableau! Great work!

                     

                    I was just thinking I should have titled my previous post as "My Brief History of Time with Tableau."

                    2 of 2 people found this helpful
                    • 7. Re: How Do You See Tableau's Canvas?
                      Simon Runc

                      Thanks Bill (That's a great title...I'd get that Trademarked!!)

                       

                      ...I feel the 'Exotic' nature of the INCLUDE/EXCLUDE 'particles' might be my un-doing! (very much the Tableau 'Graviton'!!)...If I can just find their 'Super-symmetric' 'real-world' equivalents...I might be OK!!

                      2 of 2 people found this helpful
                      • 8. Re: How Do You See Tableau's Canvas?
                        Rody Zakovich

                        Hey Simon,

                         

                        I like your premise a lot! And Bill I like what you are getting at (And thank you for the fun Puns!).

                         

                        I see things, slightly different. Mine is much more in-tangled (You might say).

                         

                        9-23-2015 1-04-37 PM.png

                         

                        I try to imagine different levels, with an infinite amount of sub-levels (Depending on the number of nested calcs). So aggregations are always calculated by the vizLoD (If any is defined), and Table Calcs build off of that. BUT Table Calcs can never loop back to the Tableau Datasource. On the other hand you have INCLUDE/EXCLUDE LoDs, they are also dependent on the vizLoD, but they can loop back to Tableau's Datasource. That is why the sit to Right, but still slightly higher than standard aggregations. Table Calcs are lower than Aggregations and Include/Exclude LoDs because they can take the properties of either (Or Both).

                         

                        Now this get's a lot more in-tangled when you have Table Calcs interacting with Aggregations. I'd imagine they would sit to the Left of Standard Table Calcs. There could also be an argument for how you address Table Calcs, At The Level, etc. But those are still bound to the vizLoD, so they might simply exist inside of that level.

                         

                        I'm with you Simon, I still haven't got this worked out 100%, BUT this line of thinking has helped me out a lot with complex problems.

                         

                        All levels exists, they live and breathe right behind my screen, I just can't see them in plain sight!

                         

                        Regards,

                        Rody

                        • 9. Re: How Do You See Tableau's Canvas?
                          Simon Runc

                          That is brilliant!!...Great work. I think the separation between calcs that can go back to the datasource and those that don't is very insightful...and not really something I'd considered before. As you say it's hard to know how to articulate the affects of VizLoD, which affect everything. This could be a kind of 'Ether' (like the Higgs field!) in which all universes operate (could even go as far as that being the space/time fabric itself!!)

                          2 of 2 people found this helpful
                          • 10. Re: How Do You See Tableau's Canvas?
                            Rody Zakovich

                            Thanks Simon,

                             

                            It is still a work in progress, but hopefully with time I can improve it.

                             

                            I don't see the vizLoD affecting Row Level Calcs or FIXED LoDs though. They are created, regardless of the ViZ, instead they are determined by the True Datasource, and then are materialized (Not to be confused with Materializing when Optimizing the Extract) into what I call "Tableau's Datasource". They are then aggregated according to the vizLoD.

                             

                            I also don't think that FIXED LoD visit the world of Aggregation. (At least not the same as the one we think of that comes after the world of vizLoD). Instead it creates it's own Aggregated World. FIXED LoDs are determined by what Fields you address in the calc itself. Then they enter the vizLoD - Aggregate World. So it is slightly different. INCLUDE/EXCLUDE doesn't have this property. They are aggregated in the vizLoD - Aggregate World and then returned to Tableau's Datasource. BUT they are still bound to the rules/properties defined.

                             

                            This "idea" has helped me out a lot when working with INCLUDE/EXCLUDE vs FIXED. I always try to remember that INCLUDE/EXCLUDE are bound to the vizLoD FIRST, then returned Row Level to Tableau's Datasource. So I still ahve to keep in mind the Rules of vizLoD - Aggregations. FIXED on the other hand, is not bound to the same rules.

                             

                            I guess I think of FIXED as [datasourceLoD Calculations] and INCLUDE/EXCLUDE as [vizLoD LoD Calculations].

                             

                            Rody

                            • 11. Re: How Do You See Tableau's Canvas?
                              Simon Runc

                              ...Yes completely agree. FIXED LoDs are aggregated, but as you say this is VizLoD independent, unlike actual Aggregated Calculations....yes I'll need another world!! and my VizLoD 'Ether' becomes another 'passport-control' gate where by some (Row Level, FIXED LoDs) are free to pass unaffected, whereas others (well all the rest!) would be affected.

                               

                              This definition between 'Materialised' (in the database sense of the word!) and 'Tableau Datasource' is an important one, and does give a very helpful perspective on things.

                               

                              I'm also coming at this, somewhat, from the angle of thinking how this can help my 'trainees' 'Think how Tableau Thinks' (I generally only get a day or 2 with them)...many of whom are coming from Excel (and are Buyers/Finance/Managers) so have almost no experience with databases and such. I don't want to confuse them...For example I often describe and show them the FIXED LoD as a SUMIFS in Excel...this is not what's happening, but telling them that it's creating an Aggregate Table and Re-joining would be met with blank expressions!! - and in reality, apart from the 'Context Filter' part, which I explain, it makes no difference to their final use of it). For helping advanced users (it is only at this level, where I really broach INCLUDE/EXCLUDE) this has helped me no end!!...and provides me with a great framework for thinking through complicated INC/EXL calcs...We just need 9-dimensional hyper-space in order to draw it all out!! (maybe Tableau v42 may release itself from our 4D world!!)

                               

                              I must say, like many of our discussions, this has really helped me clarify (and correct some I just had plain wrong!) some of these abstract concepts.

                              2 of 2 people found this helpful
                              • 12. Re: How Do You See Tableau's Canvas?
                                Jonathan Drummey

                                What Rody drew in this post Re: How Do You See Tableau's Canvas? is much like how I see Tableau, though like all models it's a (useful) simplification. For example there are computed Sets, top & conditional filters, densification, data blending, the three levels of calculation (and where they can be computed at different points in the pipeline to return different results), multiple Marks Cards, reference lines, trend lines, etc. that all make that model even more complicated.

                                 

                                Joe Mako and I have talked about this and it's one of our frustrations with Tableau. There are a number of places where Tableau has attempted to keep things in the vein of "fast/easy/beautiful" that are incomplete solutions to the whole domain of problems and as soon as we step outside of the domain that Tableau is good at (the "butter zone" as kettan has been saying lately) we can end up with huge amounts of complexity very quickly, like this thread: LOD from multiple Sources?.

                                 

                                In terms of training people, the thing that I'm always trying to hammer home for any of these steps is: what is the data set you are working with (i.e. do you understand the grain/level of detail, the relationships between fields, and relative sparseness of that data set?), what is the level of detail of the particular feature or calculation & what is it aggregating over, and what is that returning to the rest of what Tableau is doing? If someone has those fundamental bits down, then tacking on knowledge of computed Sets, LOD expressions, table calcs, etc. is pretty straightforward because thay are all variations those fundamentals.

                                 

                                Jonathan

                                2 of 2 people found this helpful
                                • 13. Re: How Do You See Tableau's Canvas?
                                  Rody Zakovich

                                  Thank you Jonathan for adding to this discussion! It is very helpful.

                                   

                                  And I very much agree that my example is a "simplification" for the same reason you have pointed out. Though Tableau's core is, as you put it "fast/easy/beautiful", it can become extremely complex, very quickly.

                                   

                                  Blending brings in another realm of complexity to my model, and I wouldn't even know where to begin adding it in. Unfortunately (But Luckily) I don't have much experience with Blending, other than helping people on the forums. Much like Simon, I always try to centralize my data though my ETL process and then serve up Tableau a single view/table/query. That way I can avoid a lot of the headaches/confusion that come with blending......

                                   

                                  Rody

                                  • 14. Re: How Do You See Tableau's Canvas?
                                    Simon Runc

                                    So true!...it's all about the Grain...I actually write 'GRANULARITY' up on the wall, so I can constantly point to it as we work through examples/problems...This isn't actually a Tableau concept, but I do spend a large amount of time going through it (not sure why it's so rarely considered/understood...my theory is 'pre-canned' reporting systems, which means people never have to really consider it)...however well worth the time, as once the penny drops their Tableau (and data in general) life becomes so much easier (and pleasurable )...and mine as a Trainer becomes that much nicer too!!

                                    2 of 2 people found this helpful
                                    1 2 Previous Next