Pick your poison | LOD or table calc?

solutions consultant manager renowned shredder thank you is everyone here for pick your poison hello these are table talks very good I'm in the right place as well so let's start with a little introduction I'm renowned shredder I'm a solutions consulting manager at tableau based out of our offices in London I have the pleasure of working in a variety of different regions as you can see on the slide from Nordics Benelux Middle East to Africa in tableau terminology if I were to group this thing I would call it the other right the beauty here however is that I have access to any season any time of the day and have access to Sun any time of the year right so it's it's it's an exciting region to be covering now when I heard that you know I was gonna do the session and it was called pick your poison the first thing I thought about is my wedding now my wife she's from Czech Republic I got married outside Prague was a beautiful venue and once you get married they have a Czech tradition where they bring in two short glasses once got water and the other one's got vodka okay of course they don't tell you which ones which right and the bride goes first right and they cheat a little bit okay so what they do is the one with vodka it's got a little residue on the side so you could tell it's a bit oily so you could sell its vodka and of course the bride goes first so my wife you know she picked vodka and I picked water if you pick vodka you're gonna be the boss in the relationship you know what happens if you pick water now why am I telling you this story the reason is the objective of a happy married life right is that you get along and you you share with each other and so on at the same time you do want that vodka right it's important right and that's exactly what I want to help you with today right and that is you can get to a number in a multiple ways right however not every single way is that is the most optimal or the best way right so let me just quickly demonstrate that to you if I go into tableau and show you a little text table right so in this particular case let's just go into the presentation mode I've derived numbers right so this is sales for central region for consumer customer segments by that I'm using superstores because everyone's already familiar with superstore so I don't really need to spend a lot of time explaining you the data set and what you see here is that all the numbers are the same right but they've been derived very differently meaning one the first column that you see is a fixed LOD expression the second one is an exclude LOD expression the third one is include LOD expression the other one that the fourth one here is just normal sum of sales and aggregated calculation and the last one is a window sum of sales okay so as I said earlier many ways to get to the same number but see what happens when I play with this filter okay so I'm gonna just deselect first class and do you see the change right so the fixed sales the fixed LOD expression the first column right the number is unchanged so it somehow didn't respect the filter selection that I just made right whereas all the other ones reflect that filter selection okay so this is just one really small example of subtle nuances right around your selection of the calculation type now I'll spend a little bit more time in terms of why it's happening and what's going on under the hood but for now let's just carry on with the agenda so the agenda that I have for you today is I'm not going to assume that you all know the basics right I am actually going to introduce you to some basics some of it I'll be honest is too basic just bear with me we'll look into what elodie expressions are and then focus on a framework and try a few examples will then go into table calques moving on to then filter order of operations right this is where we'll look under the hood into the visual pipeline of tableau and see why those filter selections affected all the other calculations but not the fixed one and then finally we will look into the performance and other considerations okay so it's it's a lot of different topics that we're gonna cover if I am running a bit late I have a tendency of rambling I'll try my best not to do that today but let's just jump in right away okay so with basics I think before you get going with any of the calculations please do spend time familiarizing yourself with your data what I mean by that is look at your data look at every single row in there and know what makes it a bit unique or different from the other rows right have that appreciation why because it's important when you're building your calculation if you simply want to just try it or if in tableau times understand what that rogue linearity is if you throw in all of your dimensions onto the canvas that's your row granularity okay now one terminology that is often used in tableau help and all the other community forms as well is aggregation but at the same time a lot of people also use the terminology and that is granularity now both of them they tend to behave in an opposite manner right so when you have just let's say sum of sales that's totally aggregated right it's the highest level of aggregation right lowest granular and as you introduce new dimensions onto the canvas guess what you are doing you're increasing the granularity and reducing the aggregation level makes sense okay great now that you know how the granularity in aggregation works before you start writing your calculation I would ask you that rather than opening up the calculation window actually write down your analytical question it helps right so what are that you're trying to achieve just write down in in plain English or any other language of your choice right and try and focus on the nouns objectives and words that actually they actually help you figure out what dimensions and and measures you should be playing with also don't like don't write a massive paragraph of an analytical question right actually try and just tackle one question at a time it makes it far easier and you'll see later on what I mean by that once you've got that try and see if you can spot some analytical grammar like per each bar and so on it helps you understand the level of aggregation you're gonna play at okay it also helps you identify the dimensions that you'll need in your calculation so once you've got that you can then go ahead and apply a framework simple right okay I'm actually going to spend time on this framework over and over again during this session I know it's it's hard to read right but hopefully when I'm actually zooming into each one of these steps it'll make a lot more sense right so what I'm proposing today if you were to take two things I want you to take away a dimension framework right so this framework will allow you to find dimensions that you need for your analytical question and the second one is a measures framework right so this will tell you what you should be doing right so if you should have a table calc or if you should have an LOD expression right so literally you just run through the framework ask yourself some questions yes no yes no and in the end it'll tell you okay - a table calc for this one okay and I'll again run you through some examples and we'll look at this board the frameworks so let's start with the first one which is LOD expressions so with already expressions we have fixed include and LOD also and exclude these are the three ones that we've got so what exactly is LOD so LOD stands for level of detail right so whenever you take your dimensions and you drop them on any one of those highlighted shelves it adds to the level of detail to the visualization that's your building okay so whenever you drop something it adds to the base LOD when you drop it in any one of this other highlight dead areas it does not add to the base level of detail right so when you drop things on tooltip you're not affecting the visible of detail if you added in filters and pages exactly the same right so typically when you're building visualization when you drag your dimensions and bring it into the canvas right whatever that you see is your visible of detail okay now what LOD expressions allow you to do is they allow you to define the calculation level of detail right so this is important what I'm gonna be building is called a calculation level of detail whereas the visualization is the viz level of detail okay so the LOD expressions they allow you to define your calculation level of detail independent of what chart you're building okay and that's actually very very powerful and you'll see why so let's look at the first one which is fixed level of details what fixed level of detail allows you to do is do your calculation regardless of what your biz LOD looks like okay so you could be doing a chart at country level but do the calculation at product and country level or something completely different and and and and that's that's okay with fixed level of detail so with fixed level of details because it's independent there are no dependencies it's actually deterministic so what I mean by that is you can count on it to give you the same result over and over again and because of that you can actually use it as a dimension or a measure okay so why is it important because if it's a dimension right you could do all the cool things you can with a dimension right so you can do it you can you can apply it as a filter and so on and it depends on the type of [Music] result it's coming up with right so if your calculation is going to give you let's a minimum order date right it's it's a date field then tableau we'll put it in a dimension field right if you are calculating calculating let's say a fixed sum of sales it's it's a number it'll drop it in measures but you can easily drag it and drop it in dimensions and now will become a dimension okay but one thing to keep in mind with fixed level of details and in the filter order of operations later on in this session I'm going to talk about that a lot of times it's ignored right so when you're when you're you know making your filter selections it does not really update the results as you would expect right because it's getting calculated way before in the visual pipeline okay and then again spend time a little bit on why that's happening cool so let's start with the first application of our framework and the questions the way to analyze our questions we've been talking about so far so let's say we have a an elliptical question which is show me the sum of sales year-over-year by customer cohorts based on my customer acquisition date so in other words what I'm going for is tell me when I first acquired customer and over the lifetime what sort of sales of garden out of those cohorts okay so if our to do that little exercise right where if I just write down the question and then focus on the nouns right so all the gray ones that you see over here are our nouns right and if I were to then just look at the nouns around the prepositions right I can see that I'm looking for some of sales I'm looking for something here over here I need customer cohorts and I'm looking for a customer acquisition date right so I sort of have an idea in terms of what I should have in in in my visualization so if I look at some of sales anyone who's played with superstore we know we've got some of sales there check we don't need to figure that one out you're over here I've got our got date fields right so I can simply create an ear over ear calculation customer cohorts I don't know right I've got customer names in my data set but I don't really know what cohort they belong to and I don't know really I mean when I first acquired that customer right I would need to do some analysis on that so if I would then start mapping the fields right that I would need in my visualization for summer sales year-over-year I just needed here or a date but for customer cohorts now here's where you need to think a little bit right what am I looking for so I'm looking for a bunch of customers right so essentially what I'm going for here is a count distinct of all the customers and the customer acquisition date this is something that I don't have right so this is something that I'll have to build and what I'm literally gonna do here is just go for the first order date for that customer right in other words when is the first time they placed an order with me yeah so it's up to you I mean if you feel comfortable this way some people are just you know more talented than me they might not need they can skip that completely right but it does help me quite a lot right so now if I look at the framework how would I go about applying this let's actually you know move on to tableau and see what I'm trying to build okay so if I go into tableau and focus on this one so let's just escape out of here right so roughly here's the type of visualization I'm trying to build right so I've got ear over here right and I've got account distinct of my customers right so if i zoom in I've got count distinct of my customers and in the axis over here I'm looking for some of sales based on the cohorts right so how do we derive that well I just can simply start building this calculation right so in this case I'm using a fixed one but how did I get there right so rather than quickly showing you this calculation let me run you through that framework because that's that's that's what I wanted to share with you today right so in this one if I were to run through this framework first I would ask myself the question you know do I have all the dimensions the answer is no I do need my first order date do I need just the subset of data for this one the answer is no actually to run through all the order dates or all the rows in my data and then I move on to this piece which is do I need to assign a dimension at a row level the answer is well yeah I need to tag every single row with the the customer acquisition date right and I need to derive this customer acquisition date for each customer okay so once I follow this the answer is yes it tells me you should do a create I mean you should you should create an LED expression okay now I know it's a bit too much to follow on the on the screen at this particular point right but once you practice it a few times it'll come in quite easy so once I know the answer which is create LOD right here's where I would then come in and start building right so I need my first order date so that's my min ordered and I need to calculate it at the customer level so I plugged that in and I have my results so once I have it I can just bring that into my visualization drop it on color and that's what I was looking for so I acquired final n21 customers in 2015 and over the course of time here's their contribution and the other ones as I acquire over time okay so that's on fixed Elodie's now what's what's include I love these so with include allergies where it allows you to do it allows you to go in to a deeper level than your visible of detail okay so let's say you're operating at a country level but you want to calculate it at a state level right then include Elodie's would be would be a good one for that now earlier people applied some Jedi tricks to achieve the sort of behavior that we get with include Elodie's using index and so on we don't we don't need to do that anymore with Elodie's it makes it just much easier so let's look at one example for that okay so what I'm going for here is a data show by country count off all the products so distinct count of all the products that I sell in every single country and then I want to see off all the products that I sell in those countries which ones are unprofitable okay and I don't want to see every single product in that visualization okay so basically what I'm saying is creates a chart something like this okay where I have each one of my countries listed and the bar chart here is the distinct count of products so in France I'm selling one thousand three hundred and twenty eight different products right what I would like to see now is unprofitable products of those one thousand three and twenty eight products that I'm selling in France so let's see how would I go about doing that using the framework that we've got okay so if our to then run through this framework the important part here is the last one which is actually let me zoom in it's hard to see so the question here is does your measure aggregation need to be more granular than the VIS level of detail and the answer is yes in our case what we are really going for is calculation at the product level but visualization at the country level right so it is more granular and if and if that's the case guess what I'm gonna go with an LOD expression okay now I can spend time running through all of them right but in the interest of time I'm going to focus on the most important point where you make that decision right so how do you go about doing that well it's pretty straightforward you would just come in here and let's look at the calculation that moves bill right so you're going to use include and you're going to include at product level and then you're doing a quick conditional check right so if a product was profitable then count that product otherwise don't count that product okay so now this include calculation is gonna give me the count distinct of all the unprofitable products that I have in whatever level of detail I'm playing in right so in my case if I were to drop this in over here it'll now start working at the country level okay so once I've got this I can do whatever I want to do with it right so I can come in here and just say okay synchronize access and gives me a nice little view that okay in France two hundred thirty five products are not profitable right of the of the total 1328 that I sell over there cool so that's include with exclude the difference is that it's gonna operate in an opposite manner right so you're trying to remove things from your visible of detail okay so in this particular example I guess a good one would be if it'll stick to the country example that we've been using right I have country in my viz level of detail but the sum of sales or the percentage of sales that I'm calculating right it's at the regional level okay higher than that so I actually exclude country out of my calculation then in that case exclude Elodie's would be the right thing to do it's generally a good one for things like percentage of total difference from overall averages right now you could actually throw in totals and reference lines to achieve the same objective as well for it okay so what do include and exclude Elodie's have in common they're always a measure okay so remember I said that with fixed level of detail it could either be a dimension or a measure with include and exclude it's always a measure okay the reason for that is they're not purely deterministic they depend on the wage level of detail for their result and that's why they're always a measure they cannot be a dimension and I'll show you later on with the filter order of operations they actually are not ignored by most filters in the view so you do actually get that predictable filter behavior as you would expect out of your calculations yeah so when it comes to Elodie's now the thing is in one workbook how many views do you think you create it depends right you might ask what sort of question is that so depending on what you're trying to do you might just have couple of sheets in a workbook or you might have 200 right please don't have 200 but right you can have a lot of different views and on every single view you're trying to ask a different analytical question right and that may require you require to use a table calc LOD normal row calc aggregated calc and the list goes on now do I add every single calculation into my data model the answer is perhaps not right so this is where you create those you know in the Shelf calculation which you don't add to the data model but with Elodie's because you are operating at such different granularities right it's it's very easy for you to create a whole bunch of LOD expressions right so one at country level a continent level you know my sales region level state level zip code all of that right so I could have so many different types of calculations you can very easily overwhelm your data model with a whole list of this LOD expressions ok and if you've done it yourself it's still ok but if you share that data source with others in your business at times people might get overwhelmed with it right or they might make a mistake by dragging the wrong LOD expressions onto the view and they get a number which they were not expecting right and that could actually affect the trust in in numbers as well right so I would just say that apply some best practice around how you apply Elodie's into your data model right now I'm not gonna sit here and say like this is one way which works for everyone try and see what works for you right but what's your data model right be if you're going to share that using tableau server organize it nicely right so people need to know what what what that particular LOD expression is right so name it in a meaningful way so if you look at describe you know it's a right-click describe paste like copy that paste it in the comment so at least when people are hovering or those calculations I can see what those calculations are all about right maybe you can add a little description there which makes it quite easy as well and something that I find it quite useful is if I just organize my lod expressions and folders so you know I have separate folder for like all my fixed level of detail another folder for include and exclude so it just makes it easy for the end user who was not involved in building this data model you know it is just very easy for them to work with it ok another thing also with Elodie's is the first time you start playing with them not everyone feels comfortable with them right away ok because you're looking at something but the numbers are being calculated at a different level right it takes a bit of getting used to right so the first time when you're building I came up with this super amazing way and that was putting two views side by side have you guys done that before it's it's amazing you get to reconcile numbers it always works no but it's it's it's it's very simple right but just like you know in a dashboard what I do is I just put two charts right next to each other one with the this level of detail the way that I want it another one at the granularity that I should be calculating and then put them side-by-side okay cool enough of enough of LOD expressions let's just move on to table calyx now table calyx have been around for a long time right how many of you have been using tableau from table calculates not LOD expression days so a few of you write table caps are amazing they have a special place in my heart right Jedi's were made using table calyx right they made it look good if you're good at table calyx now what table cogs however though I do remember when people asked me freak questions had to go away do those tablecloths and then come back I had a bit of a confidence issue with table cows right because it does take a lot of practice to you know get the behavior that you're looking for now with the improvements in table calc interface it's actually become now very easy right so let me just show you a little trick that I use in building table calyx which comes in quite handy so let's go to table count building so there's a little cross tab now it's quite tempting to start building your table table calc on a visual right so let's have created a line chart or a bar chart and I won a table cat so let's say percentage difference or percentage of total or running total or so on it's quite tempting to start building it on the visualization itself I would suggest that rather than doing it on a visual do it on a text table or or a crosstab instead the answer for the reason for that is we are way comfortable with you know let's say spreadsheets you know we are very comfortable working in this environment when calculating things it's just a little bit more natural way of working or at least in my case because when I'm looking at a visual I'm not exactly sure what is across or down depending on the type of the visual that I'm looking at right so if you got a little table like this right it's very familiar right you you're used to this so let's just say that I want to start building a table calc so I've just dropped in my sum of sales I come in here and I say quick table calculation give me my percentage of total okay and I get something now what do you do next when you get this answer right you're gonna actually just scan through trying to add up things okay does this roughly add up to hundred I'm not sure so then you'll go here maybe add you know row grand totals to see oh yeah that's correct you know it's actually going this way and then adding up to hundred or depending on what you're trying to do you might a total down there to see how it's being calculated right what I would suggest is instead go to edit table calculation right and there have been new improvements right so in here it also now highlights it's called the calculation assistance right it highlights how this calculation is being performed right so depending on what I select here so if I say table down it tells me here's how its calculating if I say pain across it tells me that's the scope of my calculation it's it's fairly powerful and you don't have to close this window while you're building your table card now that previously was quite annoying because you had to close the window look at the number if it was not what you were looking for and open the window again and go back and forth right so in this case now if I want I could actually add my subtotals over here as well to see like you know how these are being calculated and so on but I'm not gonna add the subtotals okay and once you get the behavior that you're looking for you can then select specific dimensions okay because what it does that it locks down your calculation okay so it's already done the sorting for me and it's already selected that dimension that I need for my calculation so what I mean by that is once you lock it down you can change this visualization however you want right so let me actually make this standard view and say alright I want it over here great not what I want let's put it over there okay fine let's move ship moat around fantastic the thing here is though of what I'm trying to do once I've got my numbers locked down using specific dimensions it doesn't matter what I do with the visualization now my numbers will never break right so this this first-class bookcase is for furniture category this number regardless of what I try to do now will stay the same right and that's actually fairly powerful cool all right so moving on then going into our framework so let's say I'm looking for a moving average of my sum of sales on a weekly basis now Elodie's can replace table calyx in a lot of different use cases okay now one use case where the answer is always a table calc is when you need to look up previous values first values last values next values right when you're comparing across those values table calc is the only option okay so things like running totals right your your moving averages table calc is the way to go right so how do you how do you do that well pretty straightforward right so let's say this is the view that I was trying to build right where I've got my sum of sales going from one week to the other I can simply just come in here and say build my moving average right so if you know the syntax for moving average right what we are doing is we're looking at previous rows and then you can also add you know either the current row or something in future right so that's exactly what I'm doing here so I'm looking up three previous weeks and I'm also looking up a week in ahead right to then calculate my window average okay so once I got that can introduce it into my view so let's I need it as a line and then make it a dual axis okay synchronize it so Whitstable coughs you can do cool things like you know lookups right something that you're not gonna get with Elodie's now what are other use cases of table calyx so let's say I want to get average sales across regions by categories and the view that I should be building should show me both regions as well as the categories okay so if I were to show you the visualization that we're trying to build right let's say this is the sort of chart I'm going for where I'm breaking down by product categories and then I'll guard region so in this particular case this particular bar tells me that for my furniture category in central region I've got some of sales which is 470 thousand okay so what are we trying to do well we're trying to calculate average sales across regions by category right we want to show both but this should be across all regions yeah so how do I do that well if I were to run through my framework I will have my answers and the answer in here is that I need multiple marks showing the same average value okay so in other words I want to create a calculation where you know I don't care about individual region right give me average across all regions so like a little maybe you know a average line right not at this level at a pain level right so this is the sort of view that I'm going for okay so how do you do that in tableau well it's pretty straightforward right so I would just come in here and I would just say give me window average of my sum of sails that's it done right now by default what tableau is doing is actually doing it across the whole table so instead what I'm gonna do is I'm just gonna say do it for my region level okay make the size bigger perhaps what not everything just this guy right and maybe remove the borders and then make it a dual axis right so what I see now is individual region the cuts in the product category and across all regions as well the sales now as I said previously I could have easily achieved this using an average line right it's it's a matter of preference in terms of how do you want actually go about visualizing it okay cool moving on to where maybe a table calc is not the right option right so sticking to the same visualization right let's say that rather than breaking it down by region or furniture what I really want is just a high-level summary right so just my high level numbers calculate it at regional level but I don't want the view to be broken down by regions okay so this is that the high level some way sometimes I'd go on top of a dashboard right so how do I go about doing that again just write down your question run through the framework and in this particular case right if you just need summary then LOD expressions is a great way of achieving that again with table house you know you'll you need to get a bit creative with it so how do I build that calculation well I just go here to [Music] the view let's just zoom in instead of having it in the view itself I'm just saying calculate at region category using fixed LOD and give me my sum of sales so now it does not depend on my view LOD right the calculation will always give me the result that I'm looking for right so if I just drop that in here well not there on text right I get the number that I'm looking for oh cool so that's a lot of examples quite a few questions some really long framework right a lot of it you will digest you know when you actually practice right but then I think it's very important that you also developed appreciation of what's going on under the hood of tableau especially the filter mechanism because if you don't know how that operates sometimes you know your users will think that either your dashboard is broken or you know your your numbers are incorrect right so when it comes to filters in tableau we can broadly split them to to places where they get evaluated right so some of the filters get pushed back to the database and the database does all the other all the heavy lifting some of it happens locally in tableau okay so extract filters datasource context filters fixed dimension include exclude and measure filters they all happen at the database level table calc and when you do a we know right click hide that happens on local level right so it happens in tableau so let's look at some examples because without examples it is hard to wrap your head around it okay so going back into tableau I got a table again right where the first column this one is an LOD expression okay it's percentage of total sales so what this basically tells me is bookcases is twelve point three seven percent of my total sales the second column that I have is the same number but this one is a table Kalka okay so it's just been derived differently however it behaves you know quite differently depending on what you're doing right so I just deselected consumer now from here and did you notice that this one now is eleven point nine five percent and this one is twelve point three seven percent right if you didn't notice it previously earlier without the filter both of them were twelve point three seven percent right so what just happened here why did table count now suddenly is lower and the other one is higher okay why is table couch changing and the allergies are not okay something to keep in mind now other thing is however if I were to let's say move the office supplies from here then in that case both the numbers stay the same so what's going on well this first filter is your normal dimension filter so what I mean by dimension filter is I'm just bringing in a normal segment from here from my data model the dimension and then adding it to the filter shelf that's it it's a normal dimension filter yeah whereas the other one is a table calc filter right now what is the table calc filter so if I go to edit it's a look-up on itself okay so I'm doing the lookup of the category and it operates more like a hide okay it's not necessarily a filter so if I to assign the numbers here first my Elodie's are being calculated right so this is when when I was messing around with the dimension filters right my table calc numbers changed right so when I was changing this segment my table cut numbers change because it happens down the pipeline right but my fixed or the LOD expression those those results didn't change right now when I however started changing my table calc filters right it's happening all the way down in the pipeline right it just managed to hide the previous things so that's why the numbers were the same does that make sense moving on then to another example right now this time it's all Elodie's right so I'll got and it's actually pretty similar to what we saw in the first example right where the first column is fixed level of detail the other two are exclude and include okay so if we focus on Central Region for consumer segment right it's Nana in 11,000 right so let's see what happens when I update the ship mode so if deselect first-class exclude and include update but fixed sales dozen so so what's going on well it's again about the filter order of operations right so if I go in here first my fixed is being calculated so because of that when I update my dimensions right using the filter selection it does not change because this was already pre-calculated right but because this gets evaluated later on right include and exclude those numbers changed make sense so that's on filter order of operations with LOD expressions if you're using fixed as I said you can convert them into dimensions and that suddenly now makes it really really powerful what I mean by that is because it's a dimension you can do all the cool things like you know you can apply it as an action filter in your dashboard right so you can pass an LOD dimension from one sheet to the other you can take any load a dimension and throw it as a quick filter and applied globally across all you know the different data sources that you have in your workbook as well right so let's have a look at what that would look like so in here I've got this particular sheet right with this ones guard number of customers and I've created order bins right so these are the customers who basically place three orders with us these are the customers who place seven orders with us right and that's the view that I've got here now the other one that I have is just the view that I created previously okay in the first example and what I get with those already dimensions is if I throw them in quick filters guess what I can apply that filter globally right so I just come in here and just say apply two worksheets and I can pick whatever worksheets are one just like a normal dimension right so when I move this quick filter right it updates the rest of the visualizations or if I were to just focus on one down below now this one is an action filter right so LEDs can be quite powerful that way cool the thing that I guess is really important to point out is if you have to use table calyx its scope it's just the sheet right you can't pass it from one sheet to the other you can't apply them in action filters and so on with Elodie's you're not limited to that scope at a sheet level all right what are some other considerations I think the really important one here is performance so when you have level of detail expressions they're part of the query that is being sent back to your database in other words it's it's a is a sub query right so if you're doing some complex LOD expressions and your database is let's just say not that performant it can have performance implications meaning it it would result in slow visualizations because of that right so if that is the case you could possibly just use table calcs right because table cows get calculated you know locally right so if you're not returning a lot of marks back a lot of values back then table caps might just be faster to do it locally in tableau rather than doing it at the database level but on the other side then table kayaks are done locally so if you throw a lot of data locally then again that could make it slow as well so if that is the case then we'll do Elodie's instead what really helps with table caps have come to realize is if you use aggregated data extracts right that can actually give you some good performance benefits they're great now there are a few other considerations as well one being table calculations actually could be quite flexible right so you know in Excel wall like you have this relative referencing right and then you can also have this absolute referencing table Cox are sort of like that right with table across table down pain across pain down and depending on how I change my visualization right the table across suddenly now changes or the pain across certainly changes now that can be annoying right but really really powerful on some occasions right so I would say it's flexibility right but at the same time if you don't care about that flexibility then Elodie's you know are a little bit more specific right where you just say I want this to be countered at product level regardless of whatever visualization I'm gonna use in also one thing which is really really important right I've showed you two times already if you are using fixed right most of the filters selections are not gonna work right so make sure that if you want to make it work you add it to the context meaning right click on that quick filter and say add to context once you add it to the context then then the calculations will work fine right so if I were to actually just demonstrate that to you quickly so let's look at that fixed versus include exclude example right if I take this ship mode and add that to context right this numbers changed as well now right so if I go in make whatever changes all the numbers now are the same so with fixed level of detail right calculations do make sure that you add them to context if you don't want them to be ignored fantastic so what are the takeaways from today right there is no right way unfortunately right in fact one could argue with the whole framework that I've talked about today right people do have preferences preferences can come from different reasons right so I started using tableau and then table counts were introduced so I had a lot more time playing with table Calix than Elodie's right so I have a natural tendency of going to a table cats first right rather than going to a lodi's but then I have to stop myself because you know this is why allergies were created right because table cats were not the answer for everything so what I would say is do spend time thinking about you know not just the number but also if it is the right calculation type right so the multiple ways of getting to the same number but not all of them are right try and see if the calculation framework that I've introduced today actually works for you right it's not something that I've just built in isolation right this is something that I've learned over the course of time working with the product and our customers and how they work with their data and I think it works most of the times right think about the implication of your choice right so when you're creating a whole bunch of you know lod expressions and and table calyx and so on don't just like you know keep adding them to the data model right it's about data governance as well right so you're gonna publish that data source to tableau server so other people can leverage your beautiful work including all the beautiful calculations that you've created right so think about the data model think about filtering very very important right because you don't want the users to feel that the answers are incorrect and last but not the least performance is a massive one right if you got a real really complex teller the expression which results into a massive complex query and your database is not fast guess what we'll have to worry about performance right maybe in that case you can throw the calculation framework out the window and maybe go with table caption also other thing is we are always making changes to the product right so it's always a good idea to keep abreast with all the new improvements we are making in the product right so with the table calc the new way of building table counts that I showed you right it's a new way right it's a very powerful way but if you didn't notice it then you would not appreciate you know how useful it is cool so with that I shall wrap up I know it was a lot of information today right but don't worry this is gonna be shared after the conference you can go slow you can pause you can go for a coffee come back try again hopefully that'll make a lot more sense right don't forget to fill out the survey please write and I hope you enjoyed it thank you for your time [Applause] [Music] you

Channel: Tableau Software

Views: 6,152

Rating: 4.7837839 out of 5

Keywords: Visual data, Visual analytics, Business analysis, Business analytics, Business analysis tool, Data analytics tool, Data Analytics, Analytics, Analytics platform, Cloud application, Business analytics platform, data analysis, data visualization, business dashboards, business intelligence, tableau, tableau software

Id: bLKde0pTFaI

Length: 56min 2sec (3362 seconds)

Published: Tue Oct 23 2018