I enjoyed seeing the author speak at a conference, and I wanted to love this book. As it is, I'm really glad that *things like this* can exist, pushinI enjoyed seeing the author speak at a conference, and I wanted to love this book. As it is, I'm really glad that *things like this* can exist, pushing against the standard boundaries on what a PhD thesis and scholarly work is "supposed" to be---but I'm not sure what I got out of reading *this particular* book.
Of course, like any other PhD thesis, only an expert in the field will get everything out of it, even though it's far more accessible than a usual thesis. But a big part of his point seemed to be that we should do a lot more to integrate text and images (and other senses), not just convey ideas in text with the occasional image added as an illustration... yet I didn't feel like I was really getting extra meaning from his images. Either they explicitly illustrated something the text was already saying (which traditional works can do with the usual "See Figure 1 ..."), or they made visual allusions to other works (which could be done in text too, say as footnotes).
I'm sure that's largely my own shortcoming as a reader. I hope other readers get more out of the text-image integration than I did. But I just don't know how else to digest a book like this, beyond the usual approach of "text=primary, illustrations=bonus/filler."
Maybe the comics medium just didn't do it for me. I'd love to see the analogous thing for music or sound. While I'm not that much of a visual thinker, maybe I'm more attuned to the distinctions & integration between music and lyrics.
But I can say the book inspired me to try adding many more visual elements (or even interaction? an online exploratory dataviz supplement?) to my own PhD thesis, coming up soon.
I also really liked what he says about drawing as a way to think and understand. Even if I don't get that much out of seeing other people's drawings, I can appreciate the value of sketching as I work things out myself. (See note for p.79 below.)
Favorite bits: * p.35: On specialization within science: "This narrowing of focus led to fragmentation---a cascade of individual searchlights." * p.52: "Languages are powerful tools for exploring the ever greater depths of our understanding. But for all their strengths, languages can also become traps. ... The medium we think in defines what we can see." Parts of his book sound like the Sapir-Whorf Hypothesis applied to the text-image divide. Also reminds me of a conversation on the PolicyViz podcast about dataviz as just another (facet of) language, not as objective facts. * p.57: "Every language, Hayakawa suggests, 'leaves work undone for other languages to do.'" * p.58: "A description of an image never actually represents the image. Rather, as Michael Baxandall observes, it is a representation of thinking about having seen a picture---it's already formulated in its own terms." See Patterns of Intention: On the Historical Explanation of Pictures. Maybe that's why this book's use of pictures didn't click for me. I didn't have many thoughts about the pictures here beyond what's already in the text around them. * p.76: "Lakoff and Johnson and Nunez say that our fundamental concepts do not spring from the realm of pure, disembodied reason, but are grounded in our seeing and being in the world." See Metaphors We Live By, Philosophy in the Flesh: The Embodied Mind and its Challenge to Western Thought, and Where Mathematics Come From: How the Embodied Mind Brings Mathematics into Being. Reminds me of the intro to Sachs' translation of Aristotle's Physics: A Guided Study, talking about the difference between Aristotle's careful reasoning about what our bodies can experience, vs. modern physics' careful reasoning about what our measurement instruments tell us (ignoring our immediate experience). * p.79: "Drawing, as Masami Suwa and Barbara Tversky suggest, is a means of orchestrating a conversation with yourself. Putting thoughts down allows us to step outside ourselves, and tap into our visual system and our ability to see in relation. We thus extend our thinking---distributing it between conception and perception---engaging both simultaneously. We draw not to transcribe ideas from our heads but to generate them in search of greater understanding." * p.91: "Both binding agent and action, imagination allows us to span gaps in perception." For some reason I love the idea of imagination as "binding agent." * p.95: "By stories, I don't mean only wondrous tales, but that most human of activities, the framing of experience to give it meaning." * p.135: So, by this point in the book I was getting tired of the author's push against "how it is" and telling us that we're all automatons, puppets, stuck in a rut. C'mon, dude, we have agency! And it really is worthwhile to understand how things became "how it is"---to respect history, not to ignore it while you blaze forward. So it was a pleasant surprise when, here, he brought it back together: "Emancipation, Bruno Latour writes, 'does not mean "freed from bonds" but well-attached.' The strings stay on. By identifying more threads of association, we are better able to see these attachments not as constraints but as forces to harness." See Reassembling the Social: An Introduction to Actor-Network-Theory. The images here show rock climbers roped for safety and sailing ships using ropes to direct their travel. Nice of him to foresee my frustration :) * p.136: Comparing European ocean navigation by "the detachment methods of Descartes" (reduce the 3D world's complexity using maps and instruments) vs. Pacific islanders' navigation by complete immersion in the world---not just where land is, but where the sea life or birds travel, where the pattern of waves changes, etc. See Strongman (2008). Detachment "scales well" (as tech companies today would say---maybe that ties to why Google et al. love machine learning over statistics), but it's not the only way or the best. * p.145: Wonderful small-multiples example of real-world variation among people with the "same" measurements. He outlined his own foot and that of other people who wear the same (10 1/2) shoe size. The lengths match, but the overall shapes differ a ton. "The great variance between my foot and these (and between one another), despite all being classified as the same size, illuminates my difficulty in finding shoes that fit." I should use this example in teaching intro stats classes.
Finally: * p.54: This is the only page in this book formatted (almost) like a traditional thesis---a wall of double-spaced text with occasional labeled figures/images. The endnote (p.162) says, "In the dissertation version of this work, I was required by the Office of Doctoral Studies to include a 'List of Figures' at the front of the document to refer solely to the 'figure' on this page---the page on which I most directly break the fourth wall as to what academic scholarship is supposed to look like. Their insistence upon having a list of figures to point to the sole page of text in a work made of figures quite poetically emphasized the point I was already making here."...more
Somewhat mistitled: should specify that it's mostly about ggplot2, not all R graphics. But it does have a few sections on specific kinds of graphs thaSomewhat mistitled: should specify that it's mostly about ggplot2, not all R graphics. But it does have a few sections on specific kinds of graphs that are better done in base R or lattice than in ggplot2.
If you're trying to learn ggplot2, and find yourself constantly googling for help, it'd probably help you to take 1-2 hrs to skim this book and get a holistic feel for how things work....more
[OK, it's obviously not a dataviz book as such, but was recommended to me by various dataviz experts and is indeed worth the read.]
Chapter 3, on Closu[OK, it's obviously not a dataviz book as such, but was recommended to me by various dataviz experts and is indeed worth the read.]
Chapter 3, on Closure, is my favorite. This must be why he defines comics the way he does: so that we can talk about what happens in the gaps/gutters between panels. Unlike a single image (i.e. most "real art" plus some comics like Far Side), comics i.e. "juxtaposed pictorial and other images in deliberate sequence" necessarily also contain sequenced gaps between those images. The artist must make deliberate decisions about where those gaps fall and how much content to include around them. The reader is forced to imagine what happens between the gaps, becoming an intimate participant in a distinct way from other art forms. As McCloud points out, when panel 1 shows an axe swinging and panel 2 shows a scream, it's *you the viewer* who imagines & perpetrates the murder. (In other art, say writing or painting or film, these gaps aren't inherent to the art form the way they are in McCloud's definition of comics. Similar gaps happen in the intermissions during multi-act plays, but not all plays have them, so the gap is not inherent to drama the way it is to comics.) There are many other nifty ideas here too, but the book is worth reading for Ch 3 alone.
So, what here best relates to dataviz? What can inspire new ideas or a better understanding of dataviz?
* p.57: I like his diagram of three aspects of visual arts: "reality" i.e. mimicking photorealism; "meaning" or "language" i.e. abstractions like words and numbers; and "the picture plane" i.e. respecting abstractions like shapes, lines, and colors in and of themselves. Realistic painting-art would be in the lower-left corner, abstract art in the top corner, and written art in the lower-right corner. Most comics tend to fall along the bottom line between reality and language: even when the pictures are cartoonish, they usually represent something in (imagined) reality, and are not about the drawn lines themselves. And there's usually some kind of language text along with the picture. I imagine most dataviz would fall along the right-side edge, between the picture plane (points on a scatter plot don't pretend to be anything realistic, just circles) and the language plane (the forms used have traditionally-assigned meanings just like words and numbers do).
* p.75-77: Nice use of bar-chart small multiples to summarize and analyze comics that would be hard to compare directly :) and very nice use of "visual inference" to show that the American and Japanese bar charts (and hence their comics) come from very different clusters/distributions!
* p.85-85: Good example of importance of choosing what to show; dataviz struggles with the same (show every data point? show only a global average? where's the ideal balance in between?)
* p.100-101: Weirdness of how time & space mix within and across panels---there must be something good to adapt to dataviz here, I just can't quite place it.
* p.105: Examples of comics where reader can choose direction (go up, down, left, right at will instead of strictly left-to-right then top-to-bottom) or author can play with expectations (top panels happen before bottom ones, so reaching across panels is time travel). Again, there's a germ of something here that I can't quite place.
* p.170: McCloud says each art follows 6 steps: Idea/Purpose, Form, Idiom, Structure, Craft, Surface. Often we may be attracted to the Surface first and work backwards in our appreciation (from colorful superheroes or cheesy jokes, back through Craft of doing them well, back through Structure of how they're arranged and composed, back to Idiom i.e. genre, finally back to Form i.e. comics vs writing vs singing vs ..., and Idea i.e. the "content" of the work). This has strong echoes in dataviz: "Oooh, bubble charts are pretty... Huh, it takes skill to make them well... Oh, I must learn to compose charts together in my infographic layout... Ah, there are other idioms or genres of chart that'd work better in different cases... Huh, dataviz isn't the only way to express all this... Aaaand I should remember to have my dataviz express an insight, not just be pretty."
* p.197: "Today, comics is one of the very few forms of mass communication in which individual voices still have a chance to be heard." ...that is, unlike film-making where so many other people contribute that it's almost never a lone author's work; or unlike painting, which usually isn't mass communication. This was written in 1993. Nowadays it's far easier for artists in many mediums to work as "individual voices" and self-publish online with mass distribution. But even so, comics are still one of the most notable mediums where a solo artist has such control over the product. Dataviz strikes me as similar in that respect. The *collection* of the data in the first place is usually a huge team effort. But once the data's out there for the public, a lone-wolf statistician or graphic designer can visualize it however they like, then spread it to the masses via social media or get it published in some periodicals or websites. (Frankly, sometimes I think the field would be better for it if there was *more* editorial control by editors who know what they're doing... but meanwhile it's a great place to flex your creative muscles.)...more
I teach a dataviz class for statisticians, and so I'm mostly looking for exemplary examples of what you might call statistical graphics: taking a spreI teach a dataviz class for statisticians, and so I'm mostly looking for exemplary examples of what you might call statistical graphics: taking a spreadsheet-shaped dataset, with observations in rows and variables in columns, and showing patterns/insights in that.
Some examples here were more what I'd call scientific illustration (how short people dunk, how cheetahs run fast). These were really informative and well-done, just not what I'm looking for.
As for the statistical graphics: this book has lots of beautiful design, but many of them didn't seem to give much insight. Even the introduction-writer admits it:
p.xv: "Sometimes a picture or graphic is indeed worth those 1,000 words. Sometimes a graphic is merely a replacement for those words, and sometimes it's an oversized dingbat, merely visually breaking up the blocks of text on the printed or web page."
What a great phrase: "oversized dingbat." I'll have to start using that.
I won't nitpick the dingbats included here. (Except for one: The Death Toll in Breaking Bad---WHY do people misuse the poor periodic table so? This list is not periodic, it's not an effective graphic form, it gives no insight, aaaargh! It's a beautiful poster, but why include such things in this book?)
But these were the ones I did like:
* How Common Is Your Birthday? raises more questions than it answers: Why the dramatic gap around 4th of July, and similarly-sharp jump on Valentine's Day? * How to Build a Dog: Family Ties is a nice example of loose clustering (without sharp class boundaries), informative sorting, and icons for annotation * Fifty States of Grey isn't that exceptional of a graphic---but I didn't realize Goodreads even had a data analyst team, much less data visualization staff! May be worth exploring for the next job search :) * A Campaign Map, Morphed By Money---cartograms have their weaknesses, but the ridiculously dramatic distortion for Ad Spending Per Voter is really effective. * Mapping Best Picture (item 5 here; I can't find the original, though this seems to be a precursor) is again a cartogram with very effective distortion. This also breaks one of my "rules"---a simple list or table is usually best just arranged in tabular form---but seeing the individual data points (movie names) on the cartogram really does work here. * Obama Was Not as Strong as in 2008, but Strong Enough: a novel graphic form, well explained when it could have been confusing * A National Report Card does a great job providing annotations and context---although the map itself would get the point across better as a simple scatterplot. * The 1% Next Door again breaks a "rule" by showing numbers on a map as text, not graphically; but it works here somehow. * Usain Bolt Against the Olympic Medalist Field Since 1896 is a clever case of the advice to show comparisons directly * Number of guns per 100 people: sometimes a simple Excel bar chart is all you need to make a powerful statement. * The United States redrawn as Fifty States with Equal Population is like the inverse of a cartogram: instead of keeping boundaries where they are but distorting shape, keep the land's surface as it is but change the boundaries until they match the data. Nifty idea, well executed. * Going, Going, Gone? (sadly I can't find the original) is a really great illustration of how baseball stadiums differ. Players recently signed to a new team and new stadium might hit balls that *would have been* home runs at their old stadium, but aren't in the new one. * Women as Academic Authors, 1665-2010 is the closest one here to what I'm seeking: show a large dataset informatively, show global comparisons cleanly, then let readers slice & dice for further detail as they please. Although some of the automatically-clustered subfields are pretty weird (in my own field of Probability & Statistics, *nobody* would consider "Negative binomial distribution" to be a major subcategory worth highlighting on its own). * 512 Paths to the White House is another great interactive: they *could* have forced the reader to navigate by choosing branches one at a time, but showing the global view first (with mouseover for detail) is so much more insightful.
As usual for Stephen Few, the advice is generally good but the graph examples are almost all fake data devoid of context. We can't learn anything abouAs usual for Stephen Few, the advice is generally good but the graph examples are almost all fake data devoid of context. We can't learn anything about the world from them... which is a problem when your book's goal is teaching you to learn about the world from graphs.
Also, there's considerable overlap here with his other books, talking about visual perception and other principles that apply to any dataviz (not just dashboards). That's handy if you don't plan to read any dataviz book beyond this one. Just be aware you can skim most of this if you've seen it before.
Still, there is some solid advice specifically about dashboards. More so than with other types of dataviz, a dashboard must fit all on one page/screen so you can see it at a glance. (Scrolling or switching tabs will make the dashboard useless.) And a dashboard tends to show several distinct kinds of data at once, so it's critical to organize it well with good clean layout & graphic design.
Highlights: * p.35: A good dashboard must be designed for a specific purpose. The intended user will need specific, customized info to meet their needs---you can't just make a general-purpose dashboard. Also, it must fit on a single screen and be designed to be monitored at a glance. * p.41: 3 classes of dashboards, by role: Strategic (for managers/execs: simple display, snapshots of long-term direction); Analytical (rich comparisons and drilldown/interactions); & Operational (dynamically show operations and flag problems in real time, allow detail on demand). * p.44: Don't just show most recent data---give context by comparing it to other times/places/competitors, or to quality standards/thresholds. * p.50: Allowing interaction to show multiple screens is OK for details-on-demand, but the main screen must show the global overview. You can't have everything fragmented with no global summary. * p.56: Excessive detail is not just unneeded---it's harmful by wasting the busy reader's cognitive load. Round down decimals and times/dates unless they are really necessary. * p.99: Dashboard should prioritize summaries (sums or averages) and exceptions (outliers, top 10 best/worst performers, etc.), to draw attention to global trend and to what needs fixing. Customize it to your audience and their goals in using this data, including the kind of summaries or charts they can be expected to read & interpret. Don't use skeumorphic displays like gauges or thermometers unless they really are necessary---usually a simple bar or line will take up less space AND allow for better comparisons. * p.121: Good to remember that sometimes we DO need a table, not a graph, e.g. for lookup: bus schedules, tax rate tables, a book's index section, etc. * p.126: Nifty idea of a "bullet graph"---basically just an annotated bar graph with a single bar, with colored background for bad/OK/good ranges and tick mark for target value. Shows the same content as the thermometers or gauges popular in dashboards, but in less space and with better usability. * p.137 and elsewhere: boo, Few uses double y-axes on some charts. This is almost always a bad idea, causing misleading interpretations (unless you are only showing one dataset on two scales, such as Celsius and Fahrenheit). * p.141: Nice example of using sparklines to give more context than a simple up/down arrow. * p.153: Good discussion of icons. Don't bother putting icons next to items that are doing well (e.g. green circle for good, red for bad). Just flag the problem cases (e.g. red circle) and leave the rest alone. This'll avoid redundancy, declutter the display, and make these problems stand out more easily to be noticed & dealt with. * p.165-8: Your design should not only support meaningful comparisons (e.g. use same color palette for your org divisions across all plots)... but also avoid meaningless ones (e.g. using the same color to mean completely different things in different plots)... and avoid implying changes when there aren't any (e.g. using thicker axes on one plot than another, without signifying anything by this change). * p.171: Most common types of dashboard interactions: drilling down for details, and slicing/subsetting the data shown. Keep interaction methods consistent throughout the dashboard. If possible, let the interaction happen on the plot directly, not through sidebar buttons etc. (which take up precious space and are far removed from the plot being modified). * p.172: I'm glad to see Few advocate for usability testing. This is hugely important. You, the designer, (1) will think some things are obvious (since you made them) that are actually hard to grasp, and (2) don't know all the details of what the subject-matter-expert user will need & do. Luckily, experience shows that just a handful of testers (5ish?) is usually enough to catch the most/worst usability problems with a design....more
I was expecting a book about dataviz specifically, but this isn't it. Here "presenting data" is meant in the sense of document design: putting togetherI was expecting a book about dataviz specifically, but this isn't it. Here "presenting data" is meant in the sense of document design: putting together reports, slideshows, and posters about your work. How to make the title page look professional? Where do you find images licensed for free use? What combinations of fonts work well for headings vs body text? and so on. There is some dataviz advice, but it's haphazard, and I disagree with plenty of it.
Evergreen seems to have plenty of solid advice on putting together reports. But since her dataviz advice is often wrong, I can't really trust her on the other matters either.
So I'm glad I skimmed this for the useful tips & resources I did find... But I cannot recommend it as a general resource to someone learning about dataviz (not sure about document design).
Good bits: * p.12: "Although working memory has limits on its cognitive load, graphic elements can reduce the overload by doing some of the thinking for the reader. By visually organizing and emphasizing information, graphic design makes it more accessible for the reader, increasing the capacity to engage with the words and data. By virtue of this process, richer chunks of information are actually created, which in turn enables the viewer to essentially handle a larger cognitive load at one time." * p.18: Apparently the International Institute for Information Design site has helpful white papers about info design. * p.44: If you use Advanced Search in Google Images, you can filter down by copyright status too, to help find images that are OK to reuse. * The APA Publication Manual apparently has helpful dataviz standards? * p.66: I didn't know how to find your fonts in Windows. Go to C:\Windows\Fonts and you can see the intended usage category for each font (text, display, decorative, etc.) * p.88: Free font websites Fontpark and Font Squirrel; and font pairing advice here and here; and an experiment on font trustworthiness * p.89: Great book title: Type & Layout: Are You Communicating or Just Making Pretty Shapes? :) * p.98: I didn't know about Adobe Kuler, a free color picker website * p.116: Not all her dataviz advice is bad: this is a good example of using diverging color schemes for a Likert scale.
Tips for next time I teach dataviz: * I like her terms "unintentional" and "sloppy." Better than my own vague explanations to students of why alignment should be perfect ("it looks almost aligned but not quite"), just say "You don't want it to look sloppy." * Show students examples of font substitution in different formats and on different machines. This is why we use PDFs and embed fonts when possible, rather than writing Word docs whose layout can get completely thrown off by using a font unavailable on the reader's machine. * Talk with students about line length: how many words or characters to fit in a block of text before making a line break? (Apparently 8-12 words, or 50-80 characters.) Useful when deciding where to put text boxes, how to shape them, how many columns to use, etc. * "Squish and separate" is graphic designers' catchier slang for the Gestalt proximity principle. * Style sheets are useful. I should review the ones I got from various newspapers when I took Alberto Cairo's dataviz MOOC.
Huh? * p.67: Do serif fonts really look that bad when projected in slideshows? Check for myself: do my Beamer slides use serifs or sans fonts? * p.72: She suggests choosing one word in the title to stand out in a decorative font (like an oldey-timey font for the word History in "KCC History Department"). I don't see the point, and it looks unprofessional to my eyes. * p.85: What exactly is wrong with bullets? * p.131: You seem to tell me to avoid putting things in the lower-left corner, because of this Gutenberg Diagram thing. But then your next example claims that putting things in that corner is a great example of the Gutenberg advice. What?
Major gripes: * p.51: No, starting bars above 0 is not "cheating a little bit"---it defeats the whole point of using bars, which is that their lengths are comparable. If you want to zoom in (and not show 0), just use dots instead of bars. Easy. And if you don't understand this, you shouldn't be in the business of doling out dataviz advice. * p.53: Weird conflicting advice: Don't use 3D because you can't read it easily against the gridlines... Instead, switch to 2D---but also remove the gridlines so that they can't be read at all...? Well, I agree 3D is bad, but this is not a coherent way to argue your point. * p.85: Oh, now you *do* include 0s, in a scatterplot where all points are far from 0? Why? And why not include a legend for the colors & shapes of these points? * p.105: Color-blindness is important. But it doesn't help readers to "illustrate" it with awful fuzzy greyscale versions of your images when your book is printed in black-and-white. * p.110 and many other places: "Go online to this book's website to see this image in color and then keep reading." Nope! That's not how reading a book works. Oh man, and also the text description of colors in her images doesn't match the images at all. Why am I still reading this? * p.112: AHA, now I get it! She does not care about effective data visualization. She is an infographic designer at heart: look at my giant, colorful "8%" without any context for whether that number higher/lower than average, or the past, or our targets, or anything. This is not about communicating data, just purely about making things "pretty."...more
Brainstorming and its variations, in three chunks: defining the problem, generating ideas, and creating form. Nice short interviews with designers atBrainstorming and its variations, in three chunks: defining the problem, generating ideas, and creating form. Nice short interviews with designers at the end. Very much focused on the process of doing graphic design, not on principles of what makes a design good.
Most of the suggestions/examples/exercises are really geared towards a kind of graphic design that *doesn't* mesh well with the kind of statistical dataviz that I do... But it's still a nice handy list of things to try when you're stuck in a rut.
I wish I'd seen this book in undergrad, when I first took a design course (for engineers). The instructors' attitude implied that we *have* to use such brainstorming tools to be creative, which was obviously hogwash. In retrospect, I'm sure they just wanted us to *practice* using these tools, while admitting that they are not the only ways to generate ideas and create form (as this book makes clear). Oh well.
p.15: "Most thinking methods involve externalizing ideas, setting them down in a form that can be seen and compared, sorted and combined, ranked and shared. Thinking doesn't happen just inside the brain. It occurs as fleeting ideas become tangible things: words, sketches, prototypes, and proposals."...more
Looks like a great introduction to graphic design thinking. Contains a ton of very concrete exercises---I can imagine learning a lot by working througLooks like a great introduction to graphic design thinking. Contains a ton of very concrete exercises---I can imagine learning a lot by working through them all. Sadly it's in the library's reference section, so I can't take it home to savor and read all the way through at the moment....more
This book is great at showing the breadth of topics and fields that fall under Information Design. (I had no idea there was such a thing as LitigationThis book is great at showing the breadth of topics and fields that fall under Information Design. (I had no idea there was such a thing as Litigation Graphics, and the Civic Design stuff is fascinating too. The interview with a Plain Language expert is adorable, as he tries to give all his responses in plain language too.) Chapters 2 and 3 introduce the process of info design from an organized perspective---it's almost more about the project-management of such tasks than about how to *do* the component steps.
However, many of the actual examples are printed too small to read the details that make the visualization useful!
So, while there are plenty of case studies, they provide very little inspiration to follow. That is, the associated interviews told me a lot about the work process, in a "day in the life of an info designer" kind of way... but the graphics themselves were just little stylistic glimpses, not big enough to appreciate how the *information* is *designed* to be *useful* (which I thought was the point of the book!)
So, I might recommend this if you're considering hiring an info design expert to help your organization... or if you're about to start your own info design company and already have the core skills, just not the project-management experience... but not if you want to learn the basics of doing info design work yourself.
Ideas for what to tell my Data Visualization students: * Consider the device type (desktop, laptop, smartphone) your readers will use to access your online infographic * Consider the audience: demographics (age, eyesight?), language (only one, or several? and if many, then one-at-a-time or all-at-once?), frequency of use (reading this info once, or using as a frequent reference?), etc. * Assign them a project that involves audience research, personas & scenarios, wireframes, and user testing? ... and remember to iterate! * p.80, 84: research (uncited) apparently suggests that 6 or 10 people per user-subgroup should be enough to find most of the problems with your design or product...more
There's some good stuff here, but many topics are coverIt was OK, but for a beginner like myself I'd recommend The Non-Designer's Design Book instead.
There's some good stuff here, but many topics are covered too lightly to be useful. For example: "the vast majority of available fonts are not very well designed, so beware of cheap imitations." How am I supposed to do that, if I'm a novice who can't tell what makes a well-designed font, and you don't teach me?
But still, it's a handy list of things you should try to learn (possibly elsewhere) :) For example, they mention details I'd never thought about in regards to preparing artwork for professional printing, checking the contract proof, etc. And p.38-39 discuss copyright options, which is helpful.
There's a good discussion of white space: Beginners often feel they have to fill it up, but in fact it can be used to frame and group / set apart things, or to guide the eye's flow through your piece. p.42: "You could liken this to the way a garden designer might draw your eye toward a spectacular view by using an avenue of trees, or to put it another way, by using the space between the trees." (It also makes me think of a lawn between flowers/garden patches. You don't need your entire yard to be a dense thicket; just place flowers tastefully around the lawn i.e. white space.)
The projects section is particularly nice: walking you through the steps of common design projects like business cards, party invitations, flyers, etc. Some of their examples have plenty of room for missteps (graphic design is *not* all you need to know about cartography or web design, so be careful with maps or websites) but most seem pretty helpful.
At the end is a good list of "online image resources": where to find photos, clipart, background textures, vector-based maps, etc., both free and paid.
Also, this was cute: p.22: "[Computers] don't come up with the ideas for you (at least not yet anyway), so don't be fooled into thinking they're the easy option. Think of them as an expensive marker that you have to plug in!"...more
Very accessible intro for beginners. I always thought you "just have to have an eye for good design" as if it were an inborn trait... but of course thVery accessible intro for beginners. I always thought you "just have to have an eye for good design" as if it were an inborn trait... but of course there are principles you can learn, and this book strikes me as a good place to begin.
There are plenty of before-vs-after examples, which works really well for me: showing exactly how each principle can be applied.
Some of the "after" examples are still cheesy---but at least they are cleaner and more consistent than "before," and I think that's the point. This book *doesn't* teach specific design choices for conveying a specific style (elegant, minimalist, classic, or whatever). It teaches general principles, so that once you choose a style, you can convey *that style* well (even if it's a cheesy style).
An untrained person's might convey "I don't know what I'm doing / I made these choices by default." After learning these principles, your work will convey "I know enough design to make these choices deliberately" (even if the style you choose isn't one that other people might choose).
Finally, the book has the best novice-level intro to fonts/typefaces I've ever seen (not that I'm an expert). A sensible way to categorize them (finer than just serif vs. sans), helpful examples of what does & doesn't work when, and suggestions for specific nice fonts in each category. I'm also pleased that she doesn't bother wasting time on the difference between "font" vs. "typeface" (which I've heard a million times, but can never remember, and anyway seems only relevant if you're trying not to annoy OCD graphic designers).
Notes to self:
* In a way, this gives me more appreciation for the "infographics" stuff that statisticians like to make fun of: pretty posters with random numbers made big or called out somehow (but disconnected from each other, not compared on graphs like what *I* think of as "real" data visualization). It does take design skill to make a good infographic---it's just that they highlight isolated numbers in a way that's no different from highlighting isolated words or phrases. Making the "87%" pop out in "87% of statisticians make fun of infographics" is no different than making "Shakespeare" pop out in "Shakespeare was ranked the world's bestest playwright"... whereas making a data visualization that highlights the *connections* between numbers is a different skill.
* p.9: "This book is written for [...among others...] statisticians who see that numbers and stats can be arranged in a way that invites reading rather than sleeping" :)
* p.13: Her 4 main design principles are Contrast, Repetition, Alignment, and Proximity (making for a memorable acronym). Repetition and Alignment are pretty straightforward. Proximity is a concept that's helping me understand why my page layouts never look good. I always thought it's most logical to space all the elements apart evenly... But that leaves the reader with no groupings to connect related things, and no way to tell what's important if everything is its own separate piece. See p.20-22 for a great example. Contrast seems especially relevant to dataviz too: "If the elements [...] are not the same, then make them very different." E.g. if you want people to distinguish groups on your scatterplot, make the points' colors or shapes *very* different, not just subtly different, and so on.
* p.56: "Feel free to add something completely new simply for the purpose of repetition." That is, if you just have text and no graphic that ties your layout together, consider adding something just for the sake of having a motif you can repeat. Nice example on p.82 with the small triangles. Also, "It's fun and effective to pull an element out of a graphic and repeat it." Her clipart teapot (for a tea party invite) has triangles---so she makes more of them and scatters a few about the page nicely.
* p.63: "Don't be a wimp." This caught my eye a few times flipping through the book (nice case of repetition!), but I was confused... Now I see what she means: "If the two elements are sort of different, but not really, then you don't have contrast, you have conflict. ... You cannot contrast 12-point type with 14-point type. ... You cannot contrast dark brown with black. Get serious." Also a nice example on p.68: "Are the rules supposed to be two different thicknesses? Or is it a mistake?" Again, very relevant to dataviz, as well as presentation (slideshow) design. Your audience should be able to tell when something changes, and it should be clear that the change is intentional: the contrast should be dramatic. Otherwise they'll be wondering: is that font/color/size really different or am I just seeing things? Does it signify something meaningful (helping me by flagging important differences), or was it put in there at random (just confusing the reader)? This also seems relevant to her discussion of centered alignment on p.38: "The line lengths are not the same, but they are not really different. If you can't instantly tell that the type is centered, why bother?" It's an interesting justification for (almost always) using left- or right-aligned text instead of centered: to help readers see clearly that the alignment is intentional.
* p.75: Just don't use Times Roman and Arial/Helvetica. They are so common that, even to the untrained eye, they convey "I use decades-old defaults instead of thinking about what I do," which doesn't send a professional message.
* p.100 and 176: good tips on flyers and general design-process tips, which also seem especially helpful for dataviz, academic posters, slideshows, etc. Pick a focal point and really contrast it with everything else. ("...if everything is large, then nothing can really grab a reader's attention.") Make subheadings that also really contrast the body text, so readers can skim to grasp your point and decide if it's worth their attention (or find something that'll hook them), instead of being turned off by a massive wall of text. Use proximity to group sub-parts sensibly, and use repetition and alignment to help readers navigate these sub-parts easily.
* p.104: She gets into more detail about typefaces later, but here are concrete suggestions for basic heading-vs-body distinctions. Headings: a heavy black version of a sans serif, such as Eurostile, Formata, Syntax, Frutiger, or Myriad. Body: a classic oldstyle serif (Garamond, Jenson, Caslon, Minion, Palatino, or Warnock Pro Light), or a lightweight slab serif (Clarendon, Bookman, Kepler, or New Century Schoolbook).
* p.132-138: Great introduction to typefaces, in 6 categories. I knew about serif, sans serif, and other a.k.a. just plain weird fonts. Her breakdown has 3 groups of serifs (Oldstyle, Modern, Slab Serif); 1 group for Sans Serif; and 2 groups of "other" styles (Script, Decorative). Script vs Decorative seems pretty clear, and both are best used sparingly ("if the thought of reading an entire book in that font makes you wanna throw up, you can probably put it in the decorative pot.") Sans Serifs are those without serifs; straightforward enough. Also, they usually are "monoweight"---"letterforms are the same thickness all the way around" unlike most serif fonts. But a few Sans Serifs do have some thick/thin transition, which makes them more similar to Serifs (and therefore makes them worse choices if you're trying to contrast them with a Serif). Often good for headings. As for the three Serif categories: Oldstyle are classic, invisible, and usually the best for long body text. They are not monoweight---there is a moderate transition between thin and thick parts of the strokes on each letter, and if you draw a line through the thin parts of the 'o' or 'a' it'll be diagonal rather than vertical. The serifs on lowercase letters are slanted. Modern have more dramatic differences between the thick and thin parts of the strokes, and a line through the thin parts will be vertical ("vertical stress"). The serifs are thin and horizontal, not diagonal. More elegant but less readable than Oldstyle. Slab Serifs are like Moderns but thick all around: almost no difference between thick & thin, but still vertical stress and horizontal serifs. Readable and clean.
* p.158: I didn't know there was a difference between italic vs. oblique or slanted typefaces. You can take the "roman" (standard?) version of the typeface and just slant everything, but the "italic" version is entirely redrawn---some of the letters look substantially different than just slanted versions of the roman ones. If you're using italics for contrast, make it a real contrast by using actual italics vs. roman, and not just regular vs. slanted roman.
* p.170: "Try to verbalize what you see. If you can put the dynamics of the relationship into words, you have power over it. ... Name the problem, then you can create the solution." Again, very relevant to learning dataviz by critique as well.
* p.179: "As a college teacher, all the quizzes, tests, and projects I give are 'open book, open mouth.' Students can always use their notes, they can use their books, they can talk with each other, they can talk with me. ... I was much more likely to retain the correct information if I wrote down the correct information. Rather than guessing and then writing down a wrong answer, the process of finding the correct answer on a test was much more productive." Might be worth trying next time I teach dataviz. ...more
Parts are worth the read, but other parts *really* need to be skimmed.
I was expecting, and got, a book full of interesting tidbits about the history oParts are worth the read, but other parts *really* need to be skimmed.
I was expecting, and got, a book full of interesting tidbits about the history of maps and mapmaking, interweaved with a quest to track down a modern-day map thief. There were nice insights into how maps were always used as political and military tools, how they are secured by librarians and produced by cartographers nowadays, how the trade in old maps for decorative purposes has become a bidding war and priced out the actual historians... Lots of good stuff.
But I was not expecting this to be mashed up with the author's personal sob story, journey of self-discovery, and excessive map metaphors. * What do the author's grandfather's business failures and their effect on his mother's psyche have to do with maps and map theft? NOTHING. But we get to hear all about it anyway. * Imagine a book where every friggin' noun that comes up is compared to a map, mapmaker, or famous explorer. OK, no need to imagine. This is that book! Author, you are not just like Fremont or Columbus. We don't need to hear your delusions of grandeur. Step back and let your subjects shine. * Most of all, there is just So! Much! BS! about how "Maybe Bland [the thief] and I were not so different after all," apparently because the thief was obsessed with stealing maps and the author is obsessed with getting the story. Dude, we don't care about your quest to get the backstory. You're a friggin' journalist. That's your job. If you can't get every last juicy detail, that's OK, just admit it and move on. Don't waste pages and pages on how hard it is to do your damn job. * "I heard my voice straining at the words, 'I'm a good guy.' It struck me that I had been trying not to convince Bland but to reassure myself. And I knew that I, too, was beginning to lose my way." NO. You are just acting like an unprofessional interviewer and an unprofessional writer.
But maybe it's all just a misunderstanding: Consider writing a short magazine article, say about maps. You might throw in a cheesy paragraph at the end drawing big conclusions, excessive metaphors, and unfounded comparisons, because it's OK and even fun to do that once in an article. Now consider writing a whole book on the subject. The article went cheesy once, for maybe a maximum of 1/5th of the article. How do you extrapolate the Acceptable Cheese Limit from an article to a book? Do you go cheesy in the book once? Or do you go cheesy for 1/5th of the entire friggin' book? This author clearly, and wrongly, went with the latter. If a good editor had excised all the similes and self-discovery, it would be a really good read....more
Just started, but I love the use of simple graphs in the text. Also, the mild statistical-term shoutouts (beta distribution, etc) are nice for those oJust started, but I love the use of simple graphs in the text. Also, the mild statistical-term shoutouts (beta distribution, etc) are nice for those of us in the know, while seemingly easy to skim/skip by the laymen.
The introduction is a bit over-the-top about how much he promises *not* to be over-the-top about Big Data... But there are lots of good lines too.
p.17: "Sex appeal isn't something commonly quantified like this, so let me put it in a more familiar context: translate this plot to IQ, and you have a world where the women think 58 percent of men are brain damaged." p.21: "WEIRD research: white, educated, industrialized, rich, and democratic." I hadn't heard this term before (for psych studies carried out on convenience sample of college students) but it makes sense. Ch.3: great points about how kids today are writing MUCH more than a generation ago, despite the complaints that the internet is dumbing us down. He also mentions that people who write "u" and "i" instead of "you" and "I" on Twitter tend to already do that in other forms of writing -- people are consistent across mediums in that sense. p.69: The idiosyncratic copy-and-pasted message is convincing. I would not have guessed this is a copy-and-paste job. More power to ya. p.79: The idea of "assimilation" or "dispersion" in a social network graph (due to Backstrom & Kleinberg) is interesting, but the author seems to overcomplicate it. Maybe I misunderstood, but as far as I can tell, your marriage is "assimilated" unless (1) you have no friends in common with your spouse, or (2) ALL your friends are common -- both of you only have one group of friends. It seems clear that those are both unhealthy social situations, not just for your marriage but also otherwise. So why's this a big new insight?
[I stopped taking notes around here as I finished reading, then misplaced the book for a while, so I can't remember what most of the other highlights were.]
p.164ish: The "most typical" and "most antithetical" word lists by gender+race are both fascinating and kind of sad. It's a shame that, as far as I can see, the only few science words ("feynman" and "xkcd") are on the most-antithetical-for-black-men list. But it is funny that the most-antithetical-for-asians lists are full of misspellings. p.244: [Nice explanation of statistical precision.] "Ironically, with research like this, precision is often less appropriate than a generalization. That's why I often round findings to the nearest 5 or 10 and the words 'roughly' and 'approximately' and 'about' appear frequently in these pages. When you see in some article that '89.6 percent' of people do x, the real finding is that 'many' or 'nearly all' or 'roughly 90 percent' of them do it, it's just that the writer probably thought the decimals sounded cooler and more authoritative. The next time a scientist runs the numbers, perhaps the outcome will be 85.2 percent. The next time, maybe it's 93.4. Look out at the churning ocean and ask yourself exactly which whitecap is 'sea level.' It's a pointless exercise at best. At worst, it's a misleading one." p.245: [Funny and true.] "Data sets move through the research community like yeti---I have a bunch of interesting stuff but I can't say from where; I heard someone at Temple has tons of Amazon reviews; I think L has a scrape of Facebook." p.247: Parsons code! I remember, over a decade ago, having the brilliant idea to write a computer program that would listen to a song on the radio, or to your humming, and help you identify the song. You could even link the user to Amazon or somewhere to sell them a recording and make some money off this program. It seemed like a great project for my Signals & Systems engineering class. But I couldn't immediately think of a simple way to do it, and ended up choosing a different project, and never looked into this further. So apparently the Parsons code is an old, well-known idea, and it's how Shazam does it. I wish I'd had a bit more follow-through and tried this out then!
Overall a fun, quick read, with nice insight into how Google and Facebook and their pals deal with our data and what can be learned from it....more
p.2: "In fact, scientists and engineers become dependent on graphical representations so that, in their absence, they fail to accomplish tasks, interrp.2: "In fact, scientists and engineers become dependent on graphical representations so that, in their absence, they fail to accomplish tasks, interrupt meetings in order to fetch some representation, or at least use gestures to reproduce transient facsimile in the air."...more
There's so much overlap in all these recent data visualization books (Alberto Cairo, Stephen Few, Stephen Kosslyn, etc.) that I have trouble rememberiThere's so much overlap in all these recent data visualization books (Alberto Cairo, Stephen Few, Stephen Kosslyn, etc.) that I have trouble remembering exactly what's distinct about this one vs the others. There are also a few confusing typos that change the meaning of his (otherwise good) examples.
But it's a worthwhile quick read if datavis is an area of interest to you. And I do like the examples, such as the use of histograms as colorbars on a choropleth on p.275. Chapter 5 provides particularly nice advice on fonts/typography, annotations, data transformations, etc. Chapter 6 includes a handy chart relating questions (e.g., "What ___ is the best and worst?") to corresponding statistical concepts (max and min) and possible visuals (bar chart). I also like Yau's suggestion on p.264 to consider relatability ("...let readers see the data as it pertains to them... When you note the overall trend, you most likely looked at your own age and gender to find the corresponding probability...").