I don't know what Tableau Software is and I found it pretty amazing that when I Google it I could still not figure it out because all the results were about the stock price. I think people might be focusing too much on stock price is Googling a company's name doesn't result in seeing the company's website on the first page of results.....
Tableau is a surprisingly good desktop application for business intelligence and analytics that, IMO, grew too big for its own good. You plug it into data (spreadsheets, databases, you name it) and it makes it very simple (drag-and-drop simple) to crunch some numbers and produce interesting graphs, visualizations and analysis.
Some clever stuff they had was (early-RoR-style) guessing of data semantics based on heuristics (Column labeled "date"? probably worth grouping by months. Number pairs like -0.3242,0.12345 ? probably worth plotting on a map).
I have used trial and educational versions in the past and was always pretty impressed with the ease of use and the results. The product itself was > $1k so I never actually purchased a license.
Even though I didn't follow the strategy very closely, from what I've run into it seems lke their offerings were all over the place. I saw a local newspaper using a hosted Tableau product to display data visualizations on their pages (the kind of visualizations made by a company whose DNA is desktop apps, so kind of underwhelming). Their website is all about Gartner and enterprisey lingo.
The product itself is, well, some advanced version of Excel. This means you either sell a lot of it very cheap, or very little of it for a lot of money. Website design suggests the latter. Too bad it seems to not be working.
I'm honestly sad to see them do badly as I think the product was really innovative and it really changed my way of thinking about report building and analytics tools.
It's a pretty good product but everybody and his brother sells analytics and visualization tools, most of the companies that have been around a while sell a few different tools since they've grown by acquisition.
Tableau is more differentiated than it looks, but it doesn't look very differentiated. If Microsoft ever gets around to making many of the "smart" features in Excel actually work (i.e. csv import and automatically choosing what kind of graph to make) they will have nothing to stand on.
I'm just waiting for the street to realize that Hortonworks and their competitors also have zero moat, and if anything, a GUI interface for a Hadoop cluster is value subtraction in the age of devops.
Honestly I have never used anything similar to Tableau in the balance of "straight-forwardness" and quality of results.
My use case is: here's some data, let's take a look at it and make some decent graphs including any sort of comparative analysis.
Excel is... well, Excel. The kitchen sink bundled in it contains pivot tables. Graphs are hideous out of the box, and very limited.
Apple Numbers makes cute looking graphs, but is extremely basic.
Viz toolkits in JS make you code for stuff that should be straightforward, at different levels of abstraction (Highcharts -> D3). And I have to put stuff in SQL and do the plumbing if I need any meaningful preprocessing.
Self-proclaimed "business intelligence" tools don't really target this use case. They solve enterprise problems like connecting to MDX sources that are not there for me.
I think if you are a smaller operation and you have specific requirements that you've thought out and aren't covered well by basic tools, then getting comfortable with JS and SQL (or R or matplotlib or Jupyter or whatever) is going to be the best way.
I've spent too much time watching people spend months and tens of thousands trying to bend 12 different third party tools together when a developer could do the whole thing in a week and it would actually do what you want it to do.
Nailing down the actual requirements is the hard part of building your own dashboard, it sounds like you've done that. If your needs are stable-ish or you have consistent developer resources it's going to be hard to find a better fit than that.
This. Freelancing is the future. Hiring in-house developers, or better still, freelance programmers will ensure that you will have custom-built your maximum BI tools and with a reasonable price. Of course, finding and hiring good freelancers who are not only dedicated in their craft, but also reasonably priced is a bit involved task, but certainly doable.
That's why I was repeating the importance of him having some well thought out requirements. In situations where this is true (and it really happens sometimes), I've regularly seen good developers come in under budget and mediocre ones at least come close.
In 99% cases, that happens because your requirements change. If you hire a competent developer, the total cost of development is always a pittance compared to what you pay monthly/quarterly to a product based company.
> And I have to put stuff in SQL and do the plumbing if I need any meaningful preprocessing.
Actually, the plumbing isn't as difficult as most people make it out to be. Nowadays, its the era of abstraction and FOSS infrastructure tools like jQuery, Bootstrap and like you mentioned, Highcharts. I'm a freelancer who quite recently developed a Tableau replacement for one of my clients. This client realized that all he basically wanted from tableau was a chart and some basic data manipulation like sorting, filtering and grouping (count, sum, average, etc.). All the things needed to develop this little app already existed in the FOSS world:
1. Highcharts/jqplot for charting.
2. Twitter-Bootstrap for showing a professional front-page and UI elements.
3. jQuery and jQuery-ui for DOM manipulation, AJAX handling for SQL queries, enabling drag/drop, etc.
4. PHP/Mysql on the backend (which is needed in anycase).
As for plumbing, each of these tools is so well-documented and also a simple Google search will point to tons of StackOverflow links that happily provide an answer to any and every question you may have!
tldr; Library/Framework plumbing might seem complex initially, but for a practiced Web-Developer, its like a cake-walk!
Hey, question on how licensing for highcharts works as a freelancer. Do you have your client buy a single website license with you as the dev? Or you have a highcharts dev license? Or something else?
Nope, we went for jqplot (http://www.jqplot.com/), it is GPL. The client initially wanted to go with Highcharts, but when we found about its licensing, I started looking for a FOSS alternative and this was what I found.
rms_returns - are you interested in discussing a project I'd like to have a freelancer work on? You can reach me at http://www.simplelegal.com and my HN username.
jqplot is what I use extensively. I haven't used D3 yet, but heard that it takes a lot more initial coding to come up with even a basic line/bar chart.
Power BI is Microsoft's answer to Tableau. You can judge on how well it achieves this. There's a free (as in beer) standalone Windows desktop application, and an Azure-hosted collaboration and hosting portal.
Microsoft now has PowerBI in its Office suite that is pretty powerful. Its a Tableau competitor that I believe is a big reason that their stock is falling. It integrates with Excel, Access, etc. and a ton of different other non-Microsoft services.
I'm not sure about the Desktop app, but the web app should work in any modern browser today. Features are generally prioritized using user feedback, and this is one of the top suggestions on the ideas site:
The free version of the service is too limiting to be very interesting to us. However, it is available from Microsoft resellers as a standalone for about $10.00 per month per user with no contract.
The free desktop version is not limited. It's very powerful. The growth rate is astounding. We get significant new features every month, sometimes weekly.
If you compared PowerBI to Tableau 6 months ago (or even 3 months ago) you are out of date. Its data model capability is far superior to Tableau. The formatting and graphics capability is catching up quickly. The intuitive UX is first rate.
Zeppelin is really neat, but still quite early stage (I can't even export notebooks :( ). For day to day use & teaching I've had good results with Jupyter + Spark
I'm still amazed that csv is still so poorly implemented in excel. Especially since its such an easy fix to honor byte order marks and default delimiters. I have to assume "sorta works" is the strategy. We all know how well that worked out for internet explorer.
or very little of it for a lot of money. Website design suggests the latter
In the BI market, Tableau is very much one of the low-cost, high volume players. They've done a great job getting into lots of people's hands, the question has always been whether or not they can roll the volume of cheap(ish) single user licenses into large enterprise-type deals.
Nail on the head regarding either selling a lot cheap, or little at high price. They went for the latter as you mention. I manage software licenses at my employer, and have been put off the price of Tableau, and the inflexibility they have. We're a consulting house, and mostly use Qlik. We started using Tableau last year.
The problem for us is that as a consulting house, our staff turnover rate is high (as expected I suppose), and this doesn't work well with Tableau's named license model. We've run into troubles where we reassigned licenses after people either left projects using Tableau, or left our company altogether. They ended up forcing us to buy extra licenses, which I was unhappy with.
My stance since then has been to avoid buying more licenses. We're getting our team skilled up more on web-based viz, so we can reduce our dependence on desktop applications.
A recent client of mine dealt with this by buying dedicated laptops to run Tableau on, when you needed to use it you just tracked down one of them. Seems like there should be a rule of thumb around this--software licenses should not be so expensive or inflexible that it's cheaper to buy a dedicated $1000 piece of shareable hardware than to get one for each user.
Interesting approach! How practical is it though? I can export a tbwx or whatever the extension is, so that those who need the dashboards can use Tableau Reader. I'm however wondering if such model would work for us.
The one thing that I like about Tableau is that its documents are XML, which plays nice with git.
It seemed to work reasonably well since the typical user was only needing it an hour or two each week. All of those users needed some authoring ability so Reader wasn't a solution in this case.
Extracts are binary yes. If you're using a workbook without extracts embedded, you're working with an XML file. We do that most of the time. We only embed extracts when someone with a Tableau Reader license needs to use a dashboard. We normally keep those out of source control.
Would you consider building the visualisations yourself with something like R and Shiny? Self host the server, develop in-house, no more Tableau like problems.
I was working on a project where the client wanted to use Tableau and I pushed instead for Shiny, which is the way we went. Quite happy with the decision.
It mostly depends on client needs. We had one project where I was thinking of going d3.js and wrap it around an Electron app. Our investment in R is still not at the level I'd want, we're mostly a SAS house ...
Most of our clients are tier 1, so in some instances they already have heavy investment in Qlik, but some are willing to shell out cash for Tableau where it makes sense. I've seen senior management in clients complain about the price though at times.
> My stance since then has been to avoid buying more licenses. We're getting our team skilled up more on web-based viz, so we can reduce our dependence on desktop applications.
Anecdotal, but I've seen this happen several times with clients who previously used Tableau.
It's a bit ridiculous. We get global discounts, but even then we still end up concluding that it's expensive. We only have desktop and online licenses, server licenses are rooftop crazy.
My last gig paid high five figures for a server license, and after a year of trying to get it to work as wanted, we threw it out and wrote our own in-house web solution with open source tools.
I used to work at a company where, under duress, I had to push tableau to its limits. My company spent hundreds of thousands of dollars on it. The main use was to provide reports on the data created by the application we sold.
It would be great if it just stayed as a better version of the excel chart maker. Instead, it was seen in my company as a way to replace programmers. All of these types of applications have the same flaws:
They sell a dream that you can turn a complex task that requires experts, into a simple task that anyone can do. In reality, you either have a simple application that doesn't do much, or an extremely complex application that still doesn't offer the performance and flexibility of just using an expert.
You end up creating a programming via drag and drop application that is more complicated to learn than actual programming. You replace general programmers with useless tableau specialists.
It took the tableau experts weeks to do what a programmer could do in days when complex requirements came up. Most requirements were complex.
Tableau Server is very, very slow and requires massive resources to run.
Tableau had to run the full, unfiltered query so that it could generate filter lists with all of the possible options. This query was often too large to run.
> desktop application for business intelligence and analytics
I tried to buy the desktop app a few months ago and just got bogged down with the sales guy. They tried to sell the server products (with very high per-user pricing) when all I needed was 1 desktop licence and the free viewer.
It was difficult to work out if they were a consultancy, a service, or selling a product. I think those things are in conflict with one another and make it difficult to understand the level of commitment and risk.
Tableau is a visualization software that you can bolt on top of many different data sources, from spreadsheets to BI tools and even AWS RedShift to get awesome looking dashboards and interactive visuals.
They are/were extremely popular with many sales organizations for internal dashboarding and even some media outlets used their stuff for public websites.
Sad to see this state of things for them, as the product was true good at one point. That said, I have not touched it in 3-4 years, so not sure how competitive it is now.
The product is awesome. I made some really neat visualizations and analyses with it during an MBA project using the 15-day trial. But it's ridiculously expensive - $1000 per year with some real limitations on how many data sources you can connect.
If I could dabble with it for $50/year I would. $1000 though? That limits their market tremendously to big companies only. It's hard enough to get an $80 tool approved.
--
Looks like they give it free for students now, which is cool: https://www.tableau.com/academic/student. But after learning it, there's no path to bring the software with them when they move to actual companies, because of the cost.
Tableau is sitting on top of a data warehouse anyway. $company will be shelling out much more on the data warehousing and ETL side than on the vis layer.
Even with open source tooling, it's a large and ongoing development effort to do BI right.
PowerBI may be subsidized or not. But in our use-cases and in their documentation, it really doesn't funnel customers to SQL Server. Sure, it plays nice with SQL, SSAS/MD and SSAS/Tabular. But it also plays nice with many, many other data sources individually and in mashup. Its internal data modeling and ETL capabilities rival the power of SSAS and SSIS and they are free -- I see them as more of a competitor to SQL than a front end to it.
You may, but that's not how the strategy guys at Microsoft see it, according to their own words. Data modeling and ETL capabilities rival SSAS and SSIS only in very trivial scenarios.
I'm not saying that to bash on PowerBI. I love that product and use it heavily and promote it to clients. It just serves different purposes than Tableau.
The data modeling engine IS SSAS. The only major missing pieces are row-level security and defined KPIs. The query language, DAX, and the backing columnstore database are the same engine used in SSAS Tabular.
xVelocity is only a (small) part of SSAS and it has a very limited application compared to OLAP even in the upcoming 2016 edition which greatly enhances the columnar store functionality.
Tableau offers a richer end-user customization interface. Power BI offers custom visuals, but these are definitely developer-only at this point. The behavior you get from the baked in pieces is more rigidly defined than that from Tableau (monthly releases address these things one item at a time).
Out of the box mapping is miles better in Tableau.
Overall Tableau is more fully-featured than Power BI. Tableau aims to be a complete BI presentation layer. Power BI is positioned as the self-service and personalized consumption component as a part of a larger BI presentation layer. Microsoft would prefer that the entirety of the BI stack be made up of their technologies, but Power BI can consume other data sources, and its reports can be embedded in other apps, so it fits into other technologies as well.
We could go feature-by-feature and Tableau would win the majority of presentation sophistication bullets (more fine-grained control of display, filtering, interactions - richer collection of built-in visualizations).
The differentiator for Power BI is more on the self-service end. Personalized dashboards (dashboard and report are two distinct concepts in Power BI) can be trivially created from published reports. Customized reports can easily be extended from published reports and datasets. There's a strong collaboration framework based on Office 365 groups and with lessons learned from SharePoint. There is also a pretty seamless upgrade path. Power Pivot models can currently be promoted to SSAS, and the expectation is for the same to be possible with Power BI models (all the same backing database technology).
I hate to be so vague, but there's a lot to both products. I'd be happy to dive deeper into some specific cases if you've got questions.
Since you sound fairly familiar with both platforms, what's your take on if and how well MS is doing on making PowerBI functionally equivalent to Tableau? At the rate they are going do you expect they will reach parity in the foreseeable future?
I think the answer is yes if you buy into the Microsoft stack, which includes SSRS and SSAS being used in conjunction with Power BI. SQL 2016 is a big BI release for Microsoft.
Power BI as a standalone product is pretty brutally limited in terms of data volume (250MB compressed data model is the max that can be hosted in the cloud service), and is missing the extensibility and flexibility that comes from a tool like SSRS.
As a self contained product, Tableau will likely hold the lead for some time. As a platform, I think Microsoft is beyond Tableau - they cover far more of the BI spectrum (and well, especially with SSAS) than Tableau seems to ever intend to.
Those are quite big products, making feature-by-feature comparison would be quite a daunting task.
It all boils down to this, in my opinion: PowerBI is for dashboards and reporting mostly. Tableau can do dashboards, but it can also do "analysis". Uncovering insights in the data that go way beyond simple cross-filtering.
In my mind, they are different tools for different purposes.
Just because you can, doesn't mean everyone will rush to do it. This is a self-service platform. Putting the data in the hands of decision makers. There is hierarchy in how people use such software:
The user: He uses only the apparent features in the GUI. Like the grid and aggregation formulas in Excel. He learns how to use the software from other people showing him how to do stuff.
The Power User: He has deeper needs, but can only be bothered to use features 1 or 2 levels deep in the GUI. Like pivot tables, vlookups and index/match, logical operator formulas in Excel. He learns how to use the software from tutorials.
The Advanced User: He has a task and does not mind getting his hands dirty in order to fix it. Uses DAX and Cube formulas. Perhaps even Macros. He learns by googling his problem and reading documentation.
The Developer: Solves the problems at the programmatic level.
Tableau occupies a very specific spot. It is brilliant for the User who only consumes dashboards via clicking on them. No explanation needed and it is super polished. It is also powerful enough for the Advanced User who can perform relatively sophisticated analyses from the interface. Generally speaking, it is not a good fit for the Power User who doesn't have the need to justify using it. It is also not a terribly good choice for the developer because it is too restrictive and the programmatic features are not well thought out.
Qlikview costs quite a lot and I'd argue is also enterprise-grade, and doesn't have the same cost issue plaguing Tableau. Which is sad, because Tableau is probably the better product in many ways.
I am pretty sure Qlick deployments are way more expensive than Tableau ones because as far as I know it is more capable data modeling-wise. It is also nowhere near as polished.
We're using them for a client, and pretty much every client I've worked with lately is using them. I think they'll be just fine, it's still a good product.
The AWS RedShift integration was a really good move, as one of the reasons we did not adopt them at my previous company was the requirement that the data set fit in memory of the server.
With RedShift, you offload the actual crunching to the db that can handle billions of rows, and then visualize the results.
They still are popular, especially in the government contracting world. It's almost a mandatory requirement their business analysts use Tableau when creating dashboards for clients.
I find that hard to believe unless you were searching just on Google News.
Searching "tableau" in Google stills shows their website as well as a large knowledge graph on the right hand side which tells you exactly what the company is about. There's News results at the top of the page, but there's still plenty of information, including the company website, on the first page of results.
This will happen for google searches where there's a hot news story that is more likely to be what people are looking for. For spiking searches, Google automatically applies a principle called "Query Deserves Freshness" and they rank pages that are more fresh. So the brand site and wikipedia and other more authoritative explanations of the brand will be temporarily deranked.
