HN Explorer

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

54 points by datavizq ↗ HN
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.

52 comments

[ 3.0 ms ] story [ 92.3 ms ] thread
civilian
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.
jhorman
We use redash for some basic reporting. It is nice, and development is active.
gerpsh
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.
colordrops
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/
bmh100
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

chris_wot
I can second this. Qlikview is pretty awesome.

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

bmh100
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.
chris_wot
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.

doctaj
This happens to me in Spotfire any time I'm messing with calculated values. So annoying.
spaceactuary
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".
chris_wot
You can hook it up with R though.
bmh100
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

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

MrApathy
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.
hobbe80
I like periscope.io myself, although I haven't done an in-depth comparison between the current options.
bra-ket
Pentaho, Saiku, or plain D3.js on the front end with your own middle layer
MrApathy
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.

chris_wot
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.

click170
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.

thorin
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.
ecyrb
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/

gt565k
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/

panorama
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.

gt565k
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

ryanatallah
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

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

JPKab
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?
apurvadave
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.

kposehn
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 :)

dedalus
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?

jastr
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