Ask HN: Anything better than Tableau for data viz, dashboards? | Hacker News Reader

Ask HN: Anything better than Tableau for data viz, dashboards?

54 pointsdatavizq52 comments

I am in charge of implementing a web-accessible dashboard to visualise some data we are collecting on behalf of a client.

I am currently planning to use Tableau. However, my (admittedly very limited) exposure to the platform has left heavily underwhelmed. Tableau dashboards appear expensive, slow, ugly, and completely lacking in statistical tools.

Can anyone suggest a platform that can improve on any/all of these issues?

I am experienced with Postgres, JS, Ruby and Objective-C, and also with Stata and R, so I'm not afraid of some coding. But I am looking for something considerably quicker than coding the thing from scratch.

View on HN

As someone who has multiple years of experience maintaining tableau, it is laughably Ops unfriendly.

Want to change the email address that tableau sends reports to? Requires a restart of a Tableau.

Want to update the ssl certs used in tableau? That's a restart.

Want to upgrade tableau to a new veraion? Get ready to uninstall and reinstall the new version.

Of all of the servers and services that I manage, Tableau is my least favorite. However, apparently its incredibly good at what it does. I say apparently because I maintain it but I don't use it in day to day operations.

If you want to leverage R, there is a web application framework called Shiny that lets you build interactive apps from R analyses: http://shiny.rstudio.com/gallery/

If you're using more common visualizations (e.g. bar chart, line chart, scatter plot, etc) there's an excellent js library called C3 (http://c3js.org/) that wraps charts implemented in D3 with a super-simple api. I'm a huge fan.

I did a lot of test plots with several D3-based charting libs, and found weirdness with C3, such as performance degradation over time, and oddities like including the label for the data as the first entry in the array. NVD3 seemed to be the most mature and sensible of all the libs I tried. http://nvd3.org/

I can't believe I'm going to say this, but tableau can easily be the right solution.

So, if you're dealing with large datasets stored in multiple systems (like Excel + MySQL + others) Tableau can be a boon.

The ability to use many different sources of data, create calculated fields that merge/modify other fields, and then operate against them? Quite nice.

I especially am happy with how I can create larger visualizations that work across different disparate datasets from many sources.

However, it does have quite a few issues in terms of UX, usability, etc. but so far I've liked it. Your mileage may vary :)

If you want something with a lot of batteries included, extensive through JS/HTML5, and fast, look to QlikView [1]. Message me (address in profile) if you want someone to show you around the platform.

[1]: http://www.qlik.com

In my (admittedly limited) experience, you'll probably run into some of the same issues with QlikView being "expensive, ugly, and completely lacking in statistical tools".

I have deep experience in the QlikView (QV) platform, so I can address some of the points in your experience:

> expensive

QV is not free, that's for sure. You'll be spending tens of thousands of dollars for the one-time license fee, as well 20% yearly maintenance. On the other hand, you'll be saving thousands of hours of engineering effort by not reinventing the hundreds of wheels already in the platform. Don't succumb to "not invented here" syndrome.

Check out a few dashboards I designed in just a day total [1], or the vendors demos of Twitter data [2], HealthData.gov data [3], or Salesforce.com data [4].

> ugly

If you are a first time user, the default visualizations are ugly, no doubt. But for someone with design skill, the visualizations can be made quite beautiful. For wanting more tools, just add your favorite JS library and HTML to make a custom visualization.

> completely lacking in statistical tools

Fortunately, QV can link with R, allowing you to all the advanced capabilities you need. Need something more specific? Throw a microservice REST API on top of your desired application, and load that in through a GET request. There are Hadoop connectors also built in.

One thing that people often don't realize when comparing Tableau and QV, is that QV is a platform, as opposed to Tableau being just a visualization tool. QV includes ETL, task scheduling, and an in-memory analytics database.

[1]: https://imgur.com/a/3Tzni

[2]: http://us-d.demo.qlik.com/detail.aspx?appName=Social%20Media...

[3]: http://us-d.demo.qlik.com/detail.aspx?appName=Epidemiology%2...

[4]: http://us-d.demo.qlik.com/detail.aspx?appName=Salesforce.qvw

I have to agree with this assessment. Qlikview does ETL very well.

Regarding the statistics portion of your comment, QlikView has built in functions in the load script for chi squared, t and z tests (including confidence intervals), linear regression, least squares, residuals, r squared, F statistic, etc. It doesn't offer the breadth that you may get from SAS or R (especially with R's CRAN packages), but it should be sufficient for dashboard level reporting.

Can it do histograms easily? I never needed to use it to do this but I had a friend who wanted to use it for this and the process was rather horendous.

Anyone know when the next version of Qlikview is being released? Or is it going to be Qliksense all the way now?

Yes, QlikView can do histograms easily. They're basically bar charts where the expression is counted, not summed, with the width between bars set to zero. They're not labeled as histograms in QlikView, which may be why there was some confusion.

You can hook it up with R though.

I can second this. Qlikview is pretty awesome.

Except when it crashes and loses 5 hours of work, even though I was saving diligently.

Personally, I have never experienced that issue, after 4 years of working with the software. Diligent saving always takes care of the issue. There are also "Backup" and "Auto Recover" settings. You could backup the most recent 5 saves, and have Auto Recover trigger every 15 minutes. Losing 5 hours of work should never happen with all those tools to prevent it.

I did enable auto-recover but Qlikview has a tendency to bring up a prompt asking me if I want to recover files even when this wasn't necessary.

I'm very glad that YOU never experienced that issue. The issue occurred, for some reason, because Qlikview paused after a load and before it refreshed the screens and displayed a prompt that allows you to add select boxes. I thought it was ready, and I saved the file.

The fact remains that I saved diligently, and the software became unstable and lost all my work. But I guess you e never had that experience before, always fun though when some smart-arse tells you that something shouldn't happen but it does happen.

This happens to me in Spotfire any time I'm messing with calculated values. So annoying.

I like periscope.io myself, although I haven't done an in-depth comparison between the current options.

You might take a look at Bokeh (http://bokeh.pydata.org) and either the PyData stack or R (Bokeh can be used from R as well: https://github.com/bokeh/rbokeh). Bokeh inside a Jupyter notebook with widgets and/or emerging "Bokeh Apps" is a powerful application stack. Anaconda is a single download that can help you get started with all the tools (including R): http://continuum.io/downloads

It still requires some coding but it is very powerful. There are a lot of examples in the Bokeh gallery and in examples directory: https://github.com/bokeh/bokeh/tree/master/examples .

There are several devs on Bokeh mailing list eager to help and the company behind Bokeh (http://continuum.io) can provide more significant help if you need it.

I quite like BIME (http://www.bimeanalytics.com/)

Disclaimer: I work at Tableau, but I don't claim to represent them, and I'm not in tech support or sales.

Two things that are likely: If you think the statistical tooling is limited you may not be aware that Tableau offers R integration built in. So if you're comfortable in R you can probably build what you want.

https://www.tableau.com/new-features/r-integration

If you find it slow, it may be something that tech support can help out with (changing config settings, or reworking your dashboards to be more performant). You should email whoever manages your account, or hit up https://www.tableau.com/support/request and include your contact info.

You can also email me (email in profile) and I'll get back to you from my work account with the right contacts.

I've used Dashing quite a bit, for relatively simple data, demo here: http://dashingdemo.herokuapp.com/sample

official page: http://dashing.io/

It really depends on what your use case is. There are still many unknowns related to implementing a web-accessible dashboard to make a decision.

What does your client intend to do with the dashboard? How much interactivity do they want in place? Is the data real-time? What sort of advanced statistical analysis does your client want to run?

Having answers to those might help guide you toward or away from Tableau. As an everyday user of Tableau with a solid technical background, there are some things that Tableau does well and there are some that it doesn't.

I love Tableau because it allows me to join many different data sources together quickly so I can analyze the data. I can easily drag-drop and visualize my data in many different dimensions. That said, sometimes the analysis is basic and it runs slower when there's a huge dataset.

For Cyph, I initially looked into Tableau and various analytics platforms like Mixpanel, then ultimately realised that Google Analytics (which we were already using) had an events API that worked fine for our needs.

See: https://developers.google.com/analytics/devguides/collection...

And this is what our dashboard looks like: http://i.imgur.com/F8g8Hxq.png

It's fairly basic when it comes to visualisations, but thought I'd throw it out there in case it's helpful.

I don't know about built in statistical tools beyond the basic aggregates available in SQL, but we (Fullscreen) have been using Chartio and are really enjoying it. I'd say it does great for 95% of the dashboards and visualizations we need (custom d3.js for the rest) and it plays well with our data sources. Particularly coming from Redshift data I found Chartio a lot snappier than equivalent charts in Tableau, especially for large data sets.

You can either use their UI to make charts, or write pure SQL, or my preferred method of making most of the functionality in the UI and tweaking the SQL for the last few details if you need something bespoke.

Chartio looks awesome, but I'm unable to get any pricing information from their site. Care to elaborate on the cost for a small shop to leverage this for a few dozen users?

Argo (https://argo.io) is a web-based tool that enables fast, natural-language based question asking and visualization of data. It's designed to be used by a non-technical user, so you can share dashboards and visualizations with people, and they can ask their own questions.

Under the hood, Argo uses advanced search processors to turn natural language queries into SQL, optimized for visualization.

You can request a demo on their website: https://argo.io

Disclaimer: I'm a co-founder and CTO of Argo

Kibana / ElasticSearch? It's limited, but pretty and interactive, and gets you a bunch with very limited up-front work. I'm sure you can find some better demos, but here's one: http://parlement.letemps.ch/

HUE is a similar but different alternative. The "search" tab has some great demos, but appears to be down atm: http://demo.gethue.com/

Though Tableau is incredibly expensive, I haven't yet found anything better that Tableau for web dashboard and visualization. Tableau is strictly a visualization tool and not statistical analysis tool. Tableau expects you to perform all the calculation on the back-end and send it the final data for visualization. I have also used Shiny, D3, Flot, Highchart, Chartio, Spotfire, Bime, FusionCharts,Qlickview and nothing comes close to Tableau.

I'd suggest you use a JS library like HighCharts or D3JS (or both). All you need to do is format your JSON on the back-end in the correct format and throw it into the chart's configuration.

HighCharts has an amazing API, documentation, and examples.

http://www.highcharts.com/

http://d3js.org/

What's your opinion on Highcharts usage longterm? I find it's great to get something up and running, but I've found myself hitting limitations, especially when it comes to custom design. But it's possible I may just be using it ineffectively. I always assumed that one day I should port my company's charts over to D3 if we wanted to be serious about our data visualization (which makes up a big part of our site).

In other words, is it like Bootstrap in that it's a useful starting tool, but doesn't really scale well if you want to have full design control in the longrun? Do you happen to know any notable sites using Highcharts? Thanks in advance.

You can style almost anything on the chart with CSS and HTML I believe. Custom tooltip templates, etc

The API docs are just fabulous, with examples for everything :)

http://api.highcharts.com/highcharts

QlikView is similar to Tableau, though more powerful and with a steeper learning curve. But if Tableau is too expensive, likely that QlikView is, too.

Another option is Looker, a relatively new product that relies more heavily on existing transaction/DW infrastructure. Dashboards are not ugly.

You can also look at d3, though by comparison development time will be much slower than the other two I've named.

If you've got SQL experience it's actually very easy to learn Qlikview scripting language. I recently finished a gig where the CEO retrenched all the IT staff and didn't bother to have any of the Qlikview dashboard processes updated. I had to pick up Qlikview in a few days, and had it all worked out pretty fully within 2 weeks. Learning "set-analysis" took me few days to get up to speed.

It's honestly not that hard to understand. The difficulty as always is putting together a sane data model.

You might want to try out Tibco Spotfire; it is similar to Tableau and Qlik.

They do provide some statistical tools (I personally don't know much about it). And they do have a WebPlayer to view the dashboards online.

See demos here: http://spotfire.tibco.com/demos

I use Spotfire regularly, and it's a pretty good tool. I highly suggest getting professional training, though, because our team didn't and it just took FOREVER for everyone to get in the swing of things.

It's really good for allowing your end users to explore data - it just requires a lot of development time to make really usable (ie: to make it more than "just a dashboard").

It has been built with R in mind from day 1. They have their own "Tibco Enterprise Runtime for R (TERR)" which I don't get to play around with much, but it's an obvious place to start for advanced predictive stuff, machine learning, and general data manipulation. Using the "RinR" package, you can pretty much do anything that R can.

My only absolute HATE with it is the LACK OF FREE SUPPORT/Community. Even though they have a nice "tibbr" (Facebook for businesses, basically) especially for Spotfire support, it's all behind a login wall, so it's not indexed by Google at all and it's not very searchable in my experience. In my opinion, this is a fatal mistake with their entire solution. Forums are amazing. Forum posts stay around forever. Very rarely do you want "the newest" forum post. You usually want a SPECIFIC forum post - making tibbr an awful user experience for support.

Additionally, their OLD forum/community IS indexed by Google, so you'll end up at dead-ends (404s with the exact information you want, conveniently highlighted in Google just before the answer is presented). Only a few people have blogged about it in the past, and even those are usually old versions. Also, basically no one talks about it on StackExchange. So, I just find it really hard to find answers to specific questions - like you might naturally do when programming to get a problem fixed quickly.

That said, it's super flexible and might be worth a look. I have very limited experience with Tableau and Power BI, but those lacked some of the convenience features I was used to when I used them. Personally, I wish I were forced to just program all my data visualizations in R or Python, haha.

VQL is a gui for really quick analysis and plotting. It’s mostly for non-technical people, but we’ve had data science teams that use it to explore their data before breaking out IPython or Tableau.

I’m the founder of VQL. Nothing on our site yet, but happy to send a demo/instance. My email is jstrauss (then an @ sign) getvql.com

One candidate is http://redash.io/ My team has a backlog task to set it up. We use bigquery to hold onto our data, and redash can work with that.. It's open source, so it's free (except for one server) and looks good.

We use redash for some basic reporting. It is nice, and development is active.

My favourite right now is Looker. Dashboards are beautiful and ad hoc query creation is dead simple and on point. You will save yourself many "oh could you re-run this data but split it out by device type?" moments because your clients should be able to do it themselves extremely easily.

Jasperserver with jasper reports has dashboards and is much improved now with visualize js. The dashboarding is only in the free version i think and not sure how the cost compares to tableau. It's much cheaper than business objects or oracle bi publisher etc though.

https://en.wikipedia.org/wiki/FusionCharts

EDIT: This seems to be heavily used in industry as well as in Federal IT dashboard...etc. So it appears to be a good choice.

Interana (http://www.interana.com) which is a YC S12 company sounds like your best bet.

Whats your scale? how many events per day/month etc?

D3.js or even Google Charts will get the job done for most visualizations.

Pentaho, Saiku, or plain D3.js on the front end with your own middle layer

A new product you could take a look at is http://www.jut.io. It's in beta / free for anyone.

It's a streaming analytics development environment, and uses d3 for visualization. It ingests both events and metrics.

It's based on a high-level dataflow processing language that allows you to process your data flexibly (moving window analytics, anomaly detection, general statistical processing). you can build interactive apps & dashboards and control which facets users can manipulate.

aaand here's the disclosure - I work at Jut and run customer success.

Demoed Tableau at work a few years back and it was lacking in real-time visualizations. Is that still the case? If so, anyone have any recommendations?

I'm a user. I haven't found something as good.

Clicdata is user friendly and affordable. http://www.clicdata.com/

Looker. Looker is the best. It's an upfront monetary investment, and the language you most need to know is SQL, but I love it dearly.

http://www.bimeanalytics.com/

take a look at https://event-fabric.com/ we didn't officially launched the saas version but I can set up an account for you to try it for free of course.

Reply on news.ycombinator.com