Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Ask HN: Designing Useful Dashboards
35 points by pbowyer on March 8, 2015 | hide | past | favorite | 18 comments
I've set myself a side project to scrape and display car hire price data, to compare suppliers and trends.

I've never done this before and my original goal was to play & learn new skills - but the challenge has become displaying this information in a useful way (and storing it in a sensible format).

To make it useful, I'd want to see where the prices for one supplier stand in relation to the others for all cars (a table, easy). But then the change from yesterday/last week would be useful, and seeing trends. And perhaps some stats on how the suppliers differ - is one cheaper than the others in a specific sector?

What are good resources for choosing sensible ways to display and contrast data? I've come across Stephen Few and Edward Tufte, but nothing that gives me an 'Aha!' moment yet.



Stephen Few's books are the right answer for designing practical, actionable visualizations that a user will learn and use literally all the time. Edward Tufte is the right answer for designing deep, data-filled visualizations in print, and learning the principles to design new things in digital.

Given that you've "come across" the two best resources in the field, I wonder if your lack of an "Aha!" moment is because you're not answering real questions with it.

Stephen Few lists several requirements for a dashboard to be a dashboard, and two of them are timely information and the need to look at it constantly.

Even were this not a side project, I can't imagine "car hire price data" changes minute to minute, nor that you would need to be looking at this data many times an hour and make split-second decisions based on it.

That means you're not designing a dashboard: you're designing a report. And since you don't have a real use case, you're imagining things that might be useful for an unknown user with unknown needs, and that's not how reports or dashboards work. You design a chart to answer a question, not come up with questions to ask based on a chart (we're not talking about ad-hoc visualization for exploration, here).

There isn't going to be one good way to represent all the data at once. Instead, take this to be a report, and produce a chart, graph, trend, etc. for each interesting combination. Start simple, with individual numbers, and work up to complicated contrasts. It's a side project; the only user is yourself. Let that be enough and just learn about data visualization generally.


> Given that you've "come across" the two best resources in the field, I wonder if your lack of an "Aha!" moment is because you're not answering real questions with it.

Nailed it I think. It's one thing to do the technical exercise, another to try and get meaningful results when you don't know what you're looking for.

I haven't read Stephen Few's books, only newsletters and online articles. It sounds like his books are worth getting. I own the set of Edward Tufte's - I pored over them as a student and each is a work of art in itself.

> since you don't have a real use case, you're imagining things that might be useful for an unknown user with unknown needs, and that's not how reports or dashboards work. You design a chart to answer a question, not come up with questions to ask based on a chart (we're not talking about ad-hoc visualization for exploration, here).

I have got that one backwards before :)

Thank you for your fantastic reply!


Edward Tufte's web site, http://www.edwardtufte.com/tufte/ is absolutely awful. So much information overload. How is this guy a visualization design expert?


Many who are experts in an area are eunuchs, "they know how to do it".


Stephen Few's website isn't much better: http://www.perceptualedge.com/


Yes, it always baffled me, too! Especially because I have been studying from his books and they are exceptional both in terms of visual design and content!


They are mostly design experts not a front-end stuff


Here are some resources I found useful (on top of Tufte):

1. Work of Alberto Cairo: http://www.thefunctionalart.com. 2. Searching Pintrest for UI/UX, dashboard, information design related keywords. 3. Work of Ben Fry and Fathom: http://fathom.info. 4. Work of Georgia Lupi and Accurat: http://giorgialupi.net and http://www.accurat.it. 5. Infosthetics Blog: http://infosthetics.com. 6. Work of Gregor Aisch from New York Times: http://driven-by-data.net 7. Information Graphics book from Taschen: http://www.taschen.com/pages/en/catalogue/design/all/04984/f...

The last one is a huge brick with many interesting and/or beautiful examples and it also exists as iBook for about $10!

The above resources contain another set of links to great many sources of inspiration...


Before looking at arbitrary visualization patterns, have a look at some useful precedents. Truecar, for example, (https://www.truecar.com/prices-new/honda/fit-pricing/) visualizes similar data in a clean and readable way. Google Flights might be another (https://www.google.com/flights/#search;f=SFO;t=JFK;d=2015-03...).

There are some visualization patterns that might be useful to you for this exercise, like interactive small multiples (http://projects.flowingdata.com/tut/linked_small_multiples_d...) to compare suppliers' prices over time. Think a bit about the trends in the data you tease out (as you say, thinks like prices over time and comparing alternatives) and go in search of examples that try to achieve similar goals.


Don't know if this post (http://blog.hubspot.com/marketing/data-visualization-mistake...) can help, but I enjoyed it a lot. I think I can be applied to Dashboards too.


Thanks for sharing this, I found this useful.


So we have been using Geckoboard for our dashboarding needs for almost 3 years now with great success. I think we spend $19 a month and we have enterprise wide visibility to things that matter to us. 80% of the widgets we just use internal pre-built widgets, they have a way to feed in highchart data if they aren't flexible enough.

One thing that I think is important for the dashboards isn't so much the look, but more so the ability to present actionable data that stakeholders can quickly understand and react to.

In a lot of ways we also use the dashboards to hold ourselves accountable. (engineering, media and demand sides of the business). We've built a lot of software to support our organization, but the dashboards have become the most critical.

disclaimer: I'm in the performance-marketing space, so we have a lot of campaigns and moving parts to keep track of. With that said, even being a product-oriented company, I'd still use Geckoboard.

Also I'm in no way affiliated with Geckoboard, just a happy customer.



How easy is it to implement data into a template dashboard like - http://wrapbootstrap.com/preview/WB0R5L90S ??


If you have coding experience it should be very easy. Most charts will simply need data in JSON format passed to them. Personally I use Keen.io for storing and querying analytics data. You can get a live URL of an query that can be directly connected to your code. They also have a dashboard template to get you started.

https://keen.io/docs/data-visualization/


Not sure if it can be classified as a dashboard, but this one is gorgeous and interesting

https://aprilzero.com/


Which books did you look at exactly?


Edward Tufte's whole set, and Stephen Few's online newsletters and articles.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: