31 January 2018
Clients often ask us which data visualisation tool or dashboarding tool is actually the ‘best’. In this article we are going to make a thorough comparison between three dominant software packages: Power BI, Qlikview and Tableau. We can already give you a spoiler of the verdict before we begin. This isn’t an exclusive comparison and there is no out-and-out winner amongst this top 3.
For this article, we provide six criteria that are important for you. This is our pre-selection of the many considerations and not an exhaustive picture. The weighting that should be attributed to each criterion will vary drastically between organisations. We compared how the three applications add value to your data (the practical desire of the department) and focussed less on purchase conditions, for example. We approached it from the perspective of the analytics department, i.e. the user of the tool (not the administrator). For the purposes of the comparison, we treat the making of dashboards and reports as one and same. What’s more, this is just a set of three tools: there are new tools coming along virtually on a daily basis. However, they are the tools that pop up frequently in the considerations and usage of our clients. Also, these three tools, Power BI, Qlikview and Tableau, themselves change frequently too. So this comparison (January 2018) is merely a snapshot. It is based on the experiences of Cmotions with a host of different organisations, some larger, some smaller, in both profit and non-profit sectors, from municipalities to insurance companies and from retailers to banks. Finally, Cmotions does not take a position itself and is not reliant on any one of these three data visualisation tools.
This article is in two parts.
In Part 1:
In Part 2:
The first criterion by which we are assessing the three data visualisation tools is about administration and being future-proof. Or rather:
The choice for new tooling is always made within a certain context. In almost all cases, you will already have software that you are still using. Ideally you want to also integrate the new tool with your existing software (‘stack’ of software) and reap the benefits of the tools working together with one another. From an administration point of view too, it is beneficial if software is ‘in the family’. Generally speaking, it is always useful to pick something that your IT department is happy to administer. Avoid running ‘underground’ projects in which you try and assume responsibility for that management in your own department. In our experience, that simply isn’t sustainable. If you were to do that, you would have to build up knowledge of your own and you would have to get deeply involved in configuration if your tool stops working at a critical moment. But as an analytics department, you should be focussing on the function – not the administration – of the software. Therefore, make sure the software you choose is accepted by your IT department. You can do this by getting your colleagues from the administration department involved at an early stage in the selection process.
Power BI is a product by Microsoft. You can see that right away. Put simply, Power BI is actually the compilation of a whole series of add-ons developed over time for Excel. The programming language for making your own calculations (DAX) is also largely derived from good-old Excel. Besides good collaboration with plenty of other sources, there is extension integration with Microsoft SQL Server (and analytical services) and with Excel itself, of course. In Power BI, you can easily draw upon work that you have already done in these tools. Furthermore, Power BI easily connects with various different components (e.g. a database, Hadoop Insights) of the machine learning platform Microsoft Azure. It is clear to see that Microsoft takes the subject of analytics very seriously. The takeover of Revolution Analytics by Microsoft means that results of models build in the analysis software R or the programming language Python can easily be important to Power BI.
Qlikview works best as a stand-alone solution. This software package isn’t a ‘family’ of other analysis tools, and in this sense is similar to Tableau. An often-cited advantage of Qlikview is that the tool is able to quickly and visually connect together multiple files in multiple formats. ETL processes work simply with all kinds of SQL connections. In practice we mostly find that the tool can easily work with data dumps, e.g. from an SQL platform and/or with separate Excel files. One advantage is that Qlikview maintains a clear structure between these files.
It isn’t exactly a fair comparison: the giant of Microsoft against the smaller Qlikview and Tableau. When it comes to integration with other tools (NB this is nothing to do with integration with data sources), Tableau again has less to offer than Power BI. Although there is unanimous praise in the market about how Tableau and R work together. There is extensive connectivity between the functionality of the two tools. Tableau can directly draw on functions, libraries, packages and models. Calculations in Tableau dynamically invoke the R engine and feed the values back to Tableau via the Rserve package.
Software development happens at a very fast pace these days. Managers don’t want to go through a long decision-making process only to find out half way through next year that the package you chose is already outdated. They are working hard behind the screens to keep constantly renewing all three of these tools, just each in different ways.
Power BI delivers new updates on a weekly basis. There is an Ideas Page used by the community, where users can go to say what they would like. They apparently chose to release a minimum viable product (MVP) and immediately start adding new innovations to it. Sometimes these are updates that you had expected a long ago (e.g. to be able to choose the colour of the letters), but more often these are great improvements (the Matrix visual that works well in terms of functionality). Also watch out: ‘weekly releases of software’ doesn’t automatically mean you ‘get’ them straight away. In large organisations, software is managed in ‘packages’ and they can become reliant on the IT department to also distribute the new versions every week. If you administer the software yourself, you will simply update your software each week. Looking to the future: ‘Natural language querying’, where the user can ‘talk’ to his or her dashboard using Microsoft’s speech engine Cortana, really captures the imagination and is ambitious enough to inspire confidence in Power BI.
Qlik (the supplier of Qlikview) recently launched a new product (QlikSense) which works more intuitively than QlikView. This allows you to simply make flexible, interactive visualisations in order to take better decisions. Qlik has created it in response to the growing demand for a tool that anybody can use and it is therefore acting on the vision of growing data literacy. However, QlikSense isn’t seen as a replacement for Qlikview. Both tools are still in development. Considering that QlikSense is still relatively new, the platform does include all of the capabilities and functionalities of Qlikview. QlikSense is a more modern platform, and it does offer the capability to apply Self-Service Analytics in your organisation. QlikView is a Guided Analytics Tool, where the interface and reports are designed by a developer. With the capability to choose between Self-Service and Guided Analytics, Qlik offers a choice tailored to the future ambitions of your organisation. These developments are aimed first and foremost at different fields from PowerBI or Tableau, but the expansion of functionalities is expected to follow the general trends in BI.
Tableau has on average two new software releases per year and follows its own model of product improvement. Their changes are less incremental. Recent modifications include how alerts are set (for major outliers in your results), a whole set of new connectors, API’s and improvements to existing visuals. The maturity of the package (Tableau is one of the first players on the market) in any case inspires enough confidence in innovation. On top of that, Tableau already has advanced capabilities that the other tools only offer to a minimal extent: trending and forecasting. Typical ‘features’ of data use in the future – forecasting rather than only looking back.
In this choice between the three tools, on the basis of administration, it is a good idea to take a pragmatic approach. Don’t start from scratch: look at what you already have in your organisation and what fits in with it. If you can pick a tool that fits into your existing ‘stack’ – the software ‘family’ that you already have in your organisation – then they will add value to each other. Doing this can also cut costs: you can ‘fit the Power BI tool in’ whereas for Tableau and Qlikview you will have to set up a separate environment. If you go for Power BI – as a component of Microsoft, which markets dozens of other products – then you are choosing integration. Tableau and Qlikview are both rather more freestanding tools. That doesn’t say anything about how they work together with other data sources, but it doesn’t give the added value that Microsoft can give. Generally, all three tools are future-proof. However, they each have a very different interpretation of renewal (i.e. the frequency thereof). In descending order of confidence in their being future-proof, first comes Microsoft (thanks to its long history of software development and ambitions in analytics), Tableau (due to its existing maturity in the software) and then Qlikview (because of its slightly more rigid – traditional – approach to new releases).
Our second criterion is about how easily the tool conforms with the data used in your organisation, i.e.:
Obviously, you are after the tool that can ‘fit’ in your organisation as quickly as possible. For the most part we can say that connecting with your data sources isn’t the problem with any of the three tools. Generally you can distinguish between simple connections (loading a file) and more advanced connections (live – i.e. cloud – databases).
Power BI has a large set of connectors and treats every potential data connection as a viable connection; even if it is a separate file. This means you can make a simple link on your laptop and easily refresh your data (‘Refresh’ button) if you are working locally. So if your source file has been expanded with new data, and as long as you keep the structures the same, it is literal a question of one press of a button to load the new records. If you are working online, you can also retrieve data from the Microsoft OneDrive. There are plenty of more complex pre-installed connections (Microsoft, Oracle, Amazon) available which you can easily refresh via the ‘gateway’, which Power BI offers via its publication service called Power BI Service. In addition, the settings for a whole range of APIs (Facebook, SalesForce, Twitter) are all ready for you. Do bear in mind – when using it heavily – that Power BI imposes limits on live connections. If Power BI has to import more than a million rows over a live connection for your report, all you will get is an error notification. Conveniently that is actually the number of rows that are sent to Power BI after processing. So if you are in Power BI and use a query that merges more than ten million rows into one million rows, that’s fine. Separately from that, there is also a size limit of one gigabyte for your underlying dataset.
Qlikview prefers to work from simple data dumps. The tool is able to connect multiple files with ease, but if you run a data process in Qlikview, the application can become ‘too hard’ in terms of complexity and the likelihood of data processing errors. Because Qlikview is able to connect multiple data files in various data formats very well and quickly, this tool has legions of fans. There is the possibility of loading data live from an SQL platform, for example. Similarly to PowerBI, QlikView too treats every potential data connection as a viable connection and features built-in functions to facilitate the connections within the stored ETL. The size and number of files to load will determine whether the duration of the data load is longer or shorter, which largely goes unnoticed by the end user.
Power BI and Tableau</strong have been battling it out in public for over a year now. Both organisations have been making official blog posts in direct response to one another. When looking at the facts, though, Tableau presents fewer difficulties in terms of the numbers of rows and the size of datasets. You must of course bear in mind that using a two-gigabyte dataset will make using Tableau significantly slower too. Nevertheless, starting off with a large dataset is quicker and easier in Tableau. Refreshing data is comparable with the two competitors and Tableau has loads of connectors available as standard too.
Every ambition for data visualisation is unique. Therefore it is good if your tool is able to give you scope beyond standard functionality to fulfil your ambition.
The standard package of visuals in Power BI is adequate for a novice, but someone who has been using the tool intensively for two months will begin to encounter limitations. Sometimes those limitations appear with more advanced usage (complexity combined with graphs, making the X axis dynamic), sometimes even with simple tasks (until recently you could hardly even modify the text colour). For 70% of your work, the readymade visuals and perfectly satisfactory, but sometimes you want to make something a bit less generic. However, Microsoft is now more than ever drawing on an online community in which not only can users answer one another’s questions, they can also offer Custom Visuals. So not the ‘out-of-the-box’ visuals that are already in the tool as standard, but a kind of ‘after-market’ set of visuals, made by users and offered in a highly accessible way. The most eye-catching one of these is perhaps the aquarium. This video shows what is possible:
There is of course also the programming language DAX which closely resembles the language you use to write formulas in Excel. This is handy because it means you unwittingly already know a lot about it. For example, there are the famous IF and CONCATENATE functions but also many other smart ones like them (e.g. SWITCH). It is easy to create new fields (calculated columns or calculated measures), but when you start grouping and ‘binning’ things within a variable you do begin to notice that Power BI is a bit less flexible than its two competitors in our comparison. If you want to quickly group or recategorise values, it is more intuitive to do so in the other tools. It sometimes overcomplicates simple things. The same is true of handling graph labels. If you take up a notch in complexity (e.g. not only gender or age on the X axis but age groups within one gender (‘nested’)) sometimes you have to pull out some serious moves to keep it comprehensible. On the other hand, it is true that it is incredibly intuitive for filtering with its ‘slicer’. It is very easy to insert it into your worksheet. With just a couple of clicks you can sort one sub-group for your graph. On the layer beneath DAX, there is the programming language ‘M’ for even more advanced uses (e.g. scripting the loading procedures for yourself, using the Advanced Editor). Microsoft now even offers its own GitHub for developers. The emphasis in Power BI is when you want to make seriously big visualisations. For example, it you are making a chart, you can ‘only’ put 3,500 data points in it. So, in this example, you have 5,000 cities around the world and show their population sizes. The notification of the 1,500 missing data points isn’t very noticeable, so you can accidentally miss out a lot of data (and outliers!).
In Qlikview, graphs have a high standard of formatting. We can choose from lots of handy graphs and table illustrations, although they aren’t as ‘flashy’ as in its rapidly-developing competitors. It is fairly easy to make ‘Calculated Fields’, but you have to use its own language to do so. This language is a kind of variation on SQL, which is quite tricky to figure out. Qlikview is familiar with quick filters that are available in different forms (listbox, slicer). With the ability to define your tables for yourself in the web interface (FreeFrom) or the ability to send out reports in bulk (PDF), Qlikview also offers more flashy visuals for business questions, e.g. in financials, marketing or operations. The reports and dashboards are clear and don’t diverge from the purpose they have been designed for.
It terms of ease of use, Tableau is still the one that stands out. Graphics can be formatted down to the slightest detail. Unlike other tools, Tableau is able to combine graphs more easily. For example, Tableau beats its rivals on the level of detail to which you can modify a chart of projected data. When out-of-the-box visuals aren’t enough, Tableau has the solution with the Tableau Public section of its Community. Plus, Calculated Fields are easier to make and you can keep it simple when you want is dynamism. Whether you want to change a Y axis or an X axis, with a parameter and a few lines of code, into a new variable, you can make entire views dynamic. Tableau does not have a universal syntax, but anybody with a basic knowledge of programming can recognise the statements. Plus Tableau is the ideal tool for storytelling, i.e. telling a story as a ‘layer’ on top of your visuals. Animating your views is simple and it is easy to insert and format modified titles and texts. You get a lot of flexibility. After all this praise it is worth making the counterpoint that working with containers (boxes in your worksheet where your visuals appear), especially when they don’t have a fixed position and instead are ‘floating’, is unclear, and not in keeping with the otherwise user-friendly Tableau.
As long as your data sources aren’t too exclusive, you make them into a connection using your data visualisation tool. In terms of connectivity, there isn’t much between all three. Power BI only falls down because it presents limitations under large-scale usage. The degree of freedom to ‘customise’ the data does differ between the tools. Power BI delivers the basics that are sufficient for most of your dashboards. There is a growing number of options on offer to ‘customise’ things, but it is fair to say that Power BI is more for the data preparation stage than Tableau and Qlikview are. These both offer a large degree of freedom for processing details. It’s great that you can identify Power BI’s DAX language from Excel, but unfortunately you need it pretty soon if you want to make custom solutions. By the same token, Microsoft’s strategy of open-source plug-ins for Custom Visuals is more convenient than Tableau which, for example, forces you to download workbooks to edit things yourself.
Now let’s turn our attention to how each of the tools performs in two extremes of usage. One of the extremes is business intelligence – continuous control information to improve results and processes, often integrated into the organisation. The other extreme is ‘stand-alone’ data visualisation – where you can apply it to any data set by slicing and dicing separately from any other infrastructure. Let’s look at the two extremes:
Business intelligence (BI) is effectively deploying control information as a basis for making decisions. This control information is largely integrated into the organisation. This means the data source of your dashboard is part of a larger whole – the data environment of your entire organisation.
What’s in a name? Power BI draws heavily on the family of Microsoft products. There is a lot of scope for integration in the business intelligence chain. When you use Power BI it soon becomes clear that the tool is a BI-first application – and then after that also works well as a stand-alone data visualisation tool. The primary feature is data preparation. The working environment in Power BI is clearly designed with a worksheet, a data environment and an environment for links between these data sources. The conspicuous presence of the latter two environments, plus a large set of processes (including pivoting, subsets with groupings, more complex joins) go to show that ideally the user should sort the data out first before making the visualisations. It is possible to work the other way around (e.g. by having an alias for a calculated variable instead of a new field name, or stipulating that you want percentages in a visual rather than in the definition of the field) but the tool isn’t primarily designed to be used this way.
Qlikview is a stand-alone tool and can handle multiple different data sources with ease. It is possible to both set up ETL and make new variables to be displayed in a range of clear visuals. The ETL and the actual ‘app’ (with visuals) are now kept separate, in order to separate the risks from failure: if the data load fails, the app can still be kept available for the users. It is easy to set up and apply slicing and dicing in Qlikview, including using new ‘free-form’ options which generate a variety of report formats simply by clicking on variables from a pre-prepared list.
Tableau is an independent entity and has designed the software accordingly. Don’t hold your breath for much in the way of integration with Office products (e.g. to share dashboards via a shared address book). Handling more complex data models (multiple source systems, lots of tables, lots of complex ways of connecting data whilst processing it at the same time) isn’t Tableau’s trump card either. But it is our next point (standalone data visualisation) where Tableau – very deliberately – really stands out. For now. Tableau recently announced ‘Project Maestro’ – a smart data processor to be integrated with Tableau. This would appear to be a response to the broader business intelligence orientation of its competitor, Power BI.
Sometimes less interaction between different sources is required and – especially if you are an analyst – you may need simply to analyse a data set in its own right. Especially if it a data set that is ‘unrelated’ to the wider organisation. For example, if the goal is to explore correlations (data mining) in a ‘flattened’ data set.
The flip side of Power BI scoring highly on business intelligence is that is less of a ‘front end’ tool than the other two. Although it is quicker as visualising and Microsoft delivers the ‘five minutes to wow’ in terms of looks, if you use it more extensively you can sense that you need to have a solid basis first. I.e. you need to ‘knead’ the data beforehand and possibly define the data model (what keys you use to link your two sources). This is in contrast with other tools where you can do more things ‘on the fly’: at the same time as you are making the visuals. Furthermore, Power BI doesn’t really pre-select the visualisations. It leaves open all the possible visualisation options for a randomly-selected variable, even though some of the options are completely unsuitable for that case.
In Qlikview you can get straight to the nitty-gritty of stand-alone data visualisation. You can easily add new files, even unconnected ones. With its eye on the trends, Qlik wants to focus on expanding tools that are able to work with both relational and non-relational data. One first step in that direction is Qlikview’s associative abilities with data connections. As far as visuals are concerned, it offers a number of useful and clear graphs and tables. However, Qlikview doesn’t give much in the way of its own suggestions, and instead counts on the creativity of the developer. It is possible to find relationships using drill functions on aggregated data, it isn’t possible to build explanatory models on individual characteristics.
Tableau clearly wants to enable users to start working with visualisations quickly. Unlike with Power BI, you don’t need to do much ‘kneading’ of the data. Suggested graph types are cleverly pre-sorted. You can add sources in no time and you can link different types of sources (unconnected files vs. live data connection) with ease. Even without a solid basis you will still end up with decent results. Tableau is good at those quick little shortcuts that users want. Don’t bother writing out a formula yourself to categorise the values ‘Noord-Brabant’ and ‘Limburg’ as ‘Southern Netherlands’, just open a dialogue window where you can simply click on the options, redistribute them or ‘bin’ them. Using ‘Columns’ and ‘Rows’ at the top of the screen, you can conveniently make all combinations and easily integrate your visualisations. Creating filters is surprisingly more time-consuming than filtering in the other two tools, but the design is different too. A filter in Tableau can go deeper (pages) whereas slicers in Power BI are limited to the page itself. The selection function (where you can select a few data points in the graph by drawing a rectangle over the visualisation) is a very convenient feature in Tableau which Qlikview and Power BI don’t have.
Business Intelligence (BI) and stand-alone visualisation are very different – primarily in the degree of data preparation. Power BI is definitely the most demanding in terms of data preparation. Microsoft claims that it is worthwhile setting up a data engine in their ‘data-first’ way: it is ten to a hundred times faster to work with this than with Tableau. Tableau and Qlikview are certainly both faster to operate from the beginning, even if you haven’t seen the data. Therefore these three tools cover both sides of the spectrum: Power BI is BI, Tableau and Qlikview are ‘stand-alone’. Power BI as a tool with ‘back-end depth’; the other two as tools with ‘front-end depth’.
So now we have covered the first three criteria in our comparison between Power BI, Qlikview and Tableau. In Part 2:
21 February 2022
Plan your calendar free and make sure you have plenty of party clothes in your closet,... read more
7 April 2021
In these boring lockdown-times we are all desperately looking for ways to still interact with our... read more
1 April 2021
Almost every day we go online we encounter recommender systems; if you are listening to your... read more