Design checklist for perfect charts | Hacker News Reader

Design checklist for perfect charts

dashdevs.com

81 pointsDashdevs22 comments

See instead

Tufte, Edward: The Visual Display of Quantitative Information (and the followup books)

Few, Stephen: Information Dashboard Design

I'd like to add that Wall Street Journal's Guide to Informational Graphics is a great one too.

Peculiar list. Some points are good (e.g. 6, 7, 8), but overall this list has a different definition of "perfect" than expected. It's not about designing a chart that facilitates insight in the best possible way; it's about making a chart as a vehicle of parting users from their money/time (e.g. 3, 17), with particular focus on charts in mobile sites/applications.

As per 'jrd259, if you want to design chart that maximizes value provided to the user, read Tufte instead.

First item of a Design checklist: Don't use pie charts

Why?

There are multiples reasons to not use a pie/donut chart as it does not (most of the time) help to better understand the information presented. Edward Tufte and Stephen Few (both experts in data visualization) have given multiple arguments against pie charts. See the link for a good summary on the subject

https://priceonomics.com/should-you-ever-use-a-pie-chart/

You'd never use a pie chart? What if it's just more visually pleasing to the rest of your page design and you only want to convey roughly how say 3 data values compare in size?

You can use a bar chart

They may be visually pleasing, but humans just cannot process angular magnitudes at a granular level.

A bar chart where one bar is 25% larger than another is obvious, a pie chart with the same is not obvious. Numeric labels are needed, because the chart itself does not serve as a useful visual reference.

It may look nice, but it does no good. Especially once you have more than a handful of values!

My point is making sometimes look good or to draw the eye of the user can be a big enough factor to use a pie chart. Maybe you're only showing two values and only need to convey if one value is larger or a lot larger than another for example.

Not every page is a scientific document.

At this point you may as well write it longhand, "X is much larger than Y". Since promoting anything less than maximally accurate beliefs is an act of sabotage on your fellow human beings, you have to ask yourself why are you using a pie chart - because it's not to provide value for users.

That's a really extreme way to interpret what I said.

If you are comparing 2 values, sure, maybe. But 40% versus 60% still won't be the most obvious thing other than "60% is larger", understanding the magnitude of the difference will involve more thinking, flat out. Taking more mental processing power to understand a visual chart is a bug, not a feature!

My gripe with this edict is that it is not as relevant in the age of dashboards and interactive data. The same reason they were often a bad choice in static views, can be a positive within an interactive dashboard - a busy pie chart indicate visually that selection filters are not granular enough for the design criteria of the view.

Here’s the real checklist:

[ ] It is going up and to the right.

Horrible freezing on Firefox. Please fix your site.

Not sure why it loads the same recaptcha.js file 6 times or what exactly requires 200kb of CSS here, but there it is.

The amount of sarcasm in this thread already is really endearing.

"analitycs" ?

"Expences" ?

I think it's safe to say these are concepts best expressed via charts for the author.

Sarcasm aside, this is a great opportunity to bring up the work of Bret Victor again. Specifically, pushing the boundaries of how we're communicating fundamental concepts with intelligent design - it was a real eye opener for me seeing ideas that were on the verge of nonsensical in a medium (say, English) expressed effortlessly in another. I know it comes up here often but frankly it's something that everyone should be aware of.

Indeed. I recently learned of his work on HN, and have been carrying his Magic Ink paper around for 2 weeks, nibbling on it bit by bit. This was written in 2006! I see implementations of the ideas expressed everywhere - for example Netflix's "Recommended" list. Although, a nit. I once bought a bunch of fairy tale books on Amazon. For months after that I was recommended nothing but more fairy tales. It's not counter-intuitive to think that if I bought 10 things of the same kind, I may not want to buy any more of that thing for a while. I think the paper called this last-value predictor, and clearly for a subset of decisions, this predictor is maladaptive.

Olde Anglish, mayhap?

Reply on news.ycombinator.com