Skip

Bowman quits Google
March 20, 2009 10:35 AM   Subscribe

I’m thankful for the opportunity I had to work at Google. I learned more than I thought I would.... But I won’t miss a design philosophy that lives or dies strictly by the sword of data.
And with that Douglas Bowman, the great web designer and CSS guru whose hiring was considered a big coup three years ago, quits Google for Twitter.

(MeFi's own) Joe Clark approves. Bowman also confirms the "41 shades of blue" story is true -- Google can't decide on one shade of blue to use on the site, so they're experimenting with 41. (via Zeldman)
posted by dw (87 comments total) 16 users marked this as a favorite

 
Well, this is kind of inside baseball. Tim Armstrong's departure for AOL is probably a bigger deal in the scheme of things.
posted by GuyZero at 10:49 AM on March 20, 2009


When designers write about Design, and explain the hows and whys behind various decisions, I feel like I'm seeing how a magic trick is done. I worked with an accomplished ID guy for a while several years back and learned lots of little things that forever changed the way I look at things. It's one thing to have a positive visceral reaction to, say, a handtool but quite another to say why.
posted by jquinby at 10:51 AM on March 20, 2009 [5 favorites]


From his blog post:

I had a recent debate over whether a border should be 3, 4 or 5 pixels wide, and was asked to prove my case.

Engineers are asked to prove their case for adding a single byte of data to certain data structures at Google. To VPs. Yet Bowman should be allowed to do whatever he wants, anytime? Everybody is accountable for their work.
posted by GuyZero at 10:54 AM on March 20, 2009


Engineers are asked to prove their case for adding a single byte of data to certain data structures at Google. To VPs. Yet Bowman should be allowed to do whatever he wants, anytime? Everybody is accountable for their work.

First of all, I doubt that's the case with every single data structure in use at Google. Hell, all of Orkut was written ASP.NET, clearly not a carefully planned technical choice. It might be true of something that runs in the core of their mapreduce implementation or smoething.

Second of all, a single pixel doesn't actually take up any resource of google, it takes just as much bandwidth to send "border-width: 4" as it does "border-width: 5".
posted by delmoi at 10:58 AM on March 20, 2009


Also, Google's design aesthetic sucks. When they took over youtube they made it worse. Look at vimeo for an example of a video site that looks nice. Apparently Marissa Mayer is responsible for the overall design aesthetic, at least according to valleywag.
posted by delmoi at 11:00 AM on March 20, 2009


Twitter? Twitter's biggest problem is engineering! Why hire a high profile designer?
posted by mkb at 11:00 AM on March 20, 2009 [1 favorite]


Hell, all of Orkut was written ASP.NET

I'm having some trouble believing this one. Cite?

I doubt that's the case with every single data structure in use at Google

I did use the word "certain". You are correct. My point is that people are accountable for explaining their decisions. This is life in a company of nearly 20,000 people. Complaining about explaining things is like complaining about the weather - yeah it's a drag but there's simply no way around it.
posted by GuyZero at 11:03 AM on March 20, 2009


I'm having some trouble believing this one. Cite?

*ahem*
posted by mkb at 11:06 AM on March 20, 2009 [6 favorites]


Is he going to quit again when Google buys Twitter?
posted by stavrogin at 11:07 AM on March 20, 2009 [4 favorites]


Google's data-driven design process is one of its unique strengths. I can't say too much, but I personally saw how making simple design changes in the search and ads results page could result both in an objectively better user experience and millions of dollars extra revenue for Google. What's innovative is the way Google actually experiments and measures the result of those design changes, it's a much more precise form of design iteration than I'd seen before.

While that approach works well for minute changes in a relatively simple product, I think it's less successful when approaching a whole new complex product with hundreds of design elements. There you need a designer with good taste and intuition. I can imagine it being very frustrating to try to work that way in a company that thinks engineering can solve all problems. Doug's blog post is pretty clear about that.
posted by Nelson at 11:07 AM on March 20, 2009 [1 favorite]


Also, Google's design aesthetic sucks.

Typically everything gets A/B tested out the ass and they choose whatever gets the best conversion results. It's design by committee on an extreme scale. Google's not an art gallery - it's a business that's trying to maximize profits.
posted by GuyZero at 11:07 AM on March 20, 2009


GuyZero: "Engineers are asked to prove their case for adding a single byte of data to certain data structures at Google. To VPs. Yet Bowman should be allowed to do whatever he wants, anytime? Everybody is accountable for their work."

That's not his point. He didn't say that he should be able to have different rules than everyone else. Google has a method, and it has worked gangbusters. He just was sick of the method. I don't blame him.
posted by Plutor at 11:07 AM on March 20, 2009


I don't blame him.

That's because you haven't tried the free food. Although Facebook's coffee kitchens are better.
posted by GuyZero at 11:13 AM on March 20, 2009


Google is clearly over. It was a fad.
posted by bicyclefish at 11:15 AM on March 20, 2009 [1 favorite]


people are accountable for explaining their decisions

I don't know about that. Presumably Google hired him in the first place because they consider him an expert at what he does—better than whoever else was already on staff.

To be hired as an expert and be forced to justify every detail of your work to someone who is understood to know less about your job than you do? At the very least, that's a very inefficient way to work, since you would need to give everyone a design education to bring them up to speed on the principles underlying your choices. Beyond that, it would become discouraging and even insulting.

It's not surprising to hear that Google makes data-driven decisions, but it is kind of surprising that they didn't give this guy his head, when they're famous for letting employees pursue their own interests and actually deploying them.
posted by adamrice at 11:16 AM on March 20, 2009 [4 favorites]


On the one hand, yes, sometimes you need to design despite the rules. OTOH, blasting a company for looking at data when it makes decisions kinda makes my skin crawl.
posted by DU at 11:18 AM on March 20, 2009


To be hired as an expert and be forced to justify every detail of your work to someone who is understood to know less about your job than you do?

CEOs do it in every board meeting. Steve Jobs has to explain his business to a failed politician, some guy who runs a clothing store and a makeup salesperson at every board meeting. And they have the absolute power to fire him. Shockingly, he can't simply say "because I'm Steve Jobs, that's why!" every time they ask him a question.

In fairness, Steve Jobs is probably always right so questioning him is beside the point. I don't know if Bowman has a reality distortion field.
posted by GuyZero at 11:23 AM on March 20, 2009 [1 favorite]


My point is that people are accountable for explaining their decisions.

I think you're missing a key point: he wasn't asked to explain his decision, but rather to prove it. Not every decision has a proof, especially not in this field.
posted by scottreynen at 11:27 AM on March 20, 2009 [4 favorites]


GuyZero -- it might help you to actually read the post where Bowman explains his decision. He indicates that the data-driven philosophy is Google's to use, and he understands why they are going with that. And he doesn't fault them for their philosophy. As so:
I can’t fault Google for this reliance on data. And I can’t exactly point to financial failure or a shrinking number of users to prove it has done anything wrong. Billions of shareholder dollars are at stake. The company has millions of users around the world to please. That’s no easy task. Google has momentum, and its leadership found a path that works very well.
... but that's just not an environment where he would prefer to work. Classic bad match between employer and employee.

Some design chiefs like being able to work in an environment where they get to be the benevelont despot; and they should find a place that lets them do that. Doesn't necessarily mean that the shop that they're leaving is horribly dysfunctional or that the designer is overweening prima donna (though both, of course, are certainly possible. It's just not the sort of thing that is proved out by one tradenewsfilter post)
posted by bl1nk at 11:34 AM on March 20, 2009 [2 favorites]


We'll all appreciate the irony in six months or so when Google buys Twitter, right?
posted by Devils Rancher at 11:38 AM on March 20, 2009 [1 favorite]


Google has the right idea here, honestly. Their lightning-fast minimalist aesthetic ruffles the feathers of the ALA crowd? I'm glad.
posted by killdevil at 11:39 AM on March 20, 2009 [4 favorites]


Dammit, not-on-preview.
posted by Devils Rancher at 11:42 AM on March 20, 2009


but that's just not an environment where he would prefer to work. Classic bad match between employer and employee.

No, that's a fair point and I'm actually not hating on Bowman at all. There's no utopian work environment that's perfect for everyone. It is silly to micromanage anyone, especially someone that's got real chops like Bowman, but at the same time everyone gets asked to justify seemingly obvious decisions from time to time. In isolation, there's nothing wrong with being asked to explain why you chose a 5-pt border versus 3 or 4. But of course he probably didn't quit over that alone. I would just hate to see designers somehow see this as validation of the belief that testing is unnecessary or the refuge of an uncreative mind.
posted by GuyZero at 11:44 AM on March 20, 2009


he wasn't asked to explain his decision, but rather to prove it. Not every decision has a proof, especially not in this field.

How do you prove a decision? You can prove a objective claim. Was he making objective claims about his decision? If so, they should be proven. For instance, if he claims that a 5 pixel boarder is easier to see, it seems like he could cite that. If they are subjective claims (i.e. "I like it"), well they can't be proven.

If they were forcing him to make claims, I dunno.
posted by DU at 11:44 AM on March 20, 2009


Also, I did read his post, but just as everyone has their own ideas about design, I read what I read, not what you read. My reading comprehension may vary between poor and mediocre but I did indeed read the links - for those who haven't already figured it out, flaming away without RTFA is a good way to get your as handed to you on a platter. This I believe.
posted by GuyZero at 11:47 AM on March 20, 2009


I, for one wouldn't mind if Google's apps were a little easier on the eyes.
posted by !Jim at 11:49 AM on March 20, 2009 [2 favorites]


the other problem is that they can easily get stuck in a local maximum this way. By iterating through minor changes to see which one does 'best' and slowly evolving their look they may end up with a design where any minor change tests poorly, whereas a major change, or a fresh look would actually much better.

and of course there is the issue of time. A design tweak might turn some people off at first, but then, over time more people would like it. Similarly a good design gets old after a while too. So if you had two designs A, and B, and A tested better then B, you might get a positive response by shifting from A to B once people are familiar and bored with A.
posted by delmoi at 12:11 PM on March 20, 2009 [12 favorites]


One problem with A/B design is that you settle onto local optima. So you wind up testing individual, small decisions, because any large set of changes is very likely to initially test lower that the design you've been polishing for several years.

A/B testing is about meeting what the user wants now, where sometimes, design is coming up with something new and leading the user towards it.
posted by zippy at 12:18 PM on March 20, 2009 [4 favorites]


A design tweak might turn some people off at first, but then, over time more people would like it.

A lot of people HATED gmail at first. Even now... it has its faults, but so does Outlook.

So if you had two designs A, and B, and A tested better then B, you might get a positive response by shifting from A to B once people are familiar and bored with A.

Many, many companies are at the point of putting up whatever some VP chose based on their relationship with the release manager or because their dog seemed to like it. I'm sure anyone who has worked more than five minutes in web design can trot out a war story or two about bad design decisions.

A/B testing, like democracy, is the worst form of design validation except all the others that have been tried from time to time.
posted by GuyZero at 12:19 PM on March 20, 2009 [1 favorite]


Sorry, I kind of mangled my point there. Which was that while, yes, A/B testing has faults, most companies are using some technique that is a lot worse.
posted by GuyZero at 12:20 PM on March 20, 2009 [1 favorite]


CEOs do it in every board meeting. Steve Jobs has to explain his business to a failed politician, some guy who runs a clothing store and a makeup salesperson at every board meeting.

no CEOs don't and Steve Jobs doesn't have to do this with every single decision he makes. CEOs have to justify the direction they've headed to people whose money depends on his ability to steer a corporation. But Al Gore does not call Jobs to task every time Jobs turns to a subordinate and gives him something to do. Further, Gore doesn't demand proof every time Jobs claims that his way works. I guarantee you that the board of directors for Apple has confidence in Jobs' abilities.

also, you're not exactly making your point by calling a former Vice President of the united states "a failed politician," the CEO of J Crew "some guy who runs a clothing store" and the CEO of Avon "a makeup salesperson."

Again, the distinction here is how often Bowman has to not only justify, but "prove" the validity of miniscule creative decisions to people who may not even know what they're talking about.
posted by shmegegge at 12:23 PM on March 20, 2009 [1 favorite]


How do you prove a decision? You can prove an objective claim.

Actually, you can't prove a claim. And there's no such thing as an objective claim, at least not empirically speaking. Since Popper at least, most scientists who are acquainted with the philosophy of science understand that you can only falsify claims, never prove them. And so, most would agree that there are falsifiable claims, even if there aren't objective ones.

This is plainly and simply a cultural issue. I have been victimized by it in my own work. If you are dealing with an engineering culture that is used to seeing data trotted out, they will expect you to provide data. No matter if the questions the data are supposed to answer are themselves inadequate. And it certainly doesn't matter whether anything is "proven". All that matters to engineers is that some kind of supposedly measurable question is asked and a supposedly good faith effort is made to answer it that can be put in some kind of graph.

Ironically, this is an aesthetic preference on the part of engineers, and nothing more. Engineers have no interest in the substantive issue of whether some areas of expertise are more or less well addressed by graphs. They just like graphs. It comforts them to see graphs. They're like security blankets for positivists.

Designers have ways of sharing and talking about what they do that makes it easier to in fact begin to grapple with thorny aesthetic questions. But never do they pretend that focus groups or data collection are a big help in solving them. Because making aesthetic choices is about leading your audience to what they should or could want/know/feel, not asking them about what they want/know/feel already. It's about inarticulate or unexplored possibilities, not pre-existing conditions. No doubt the people at Google don't understand this and can't even begin to speak a designer's language as a result.

The fact is, it annoys the positivist engineering culture that a non-positivist culture of aesthetic judgment even exists. (It also annoyed Plato, who thought that poets had no special knowledge.) That's because engineers don't understand it, and rather than acknowledge that their way of knowing is limited and only part of the story, they would rather deny that something like valid and real aesthetic understanding even exists.

Again, it comforts them to think so.

But the world is an uncomfortable place.
posted by macross city flaneur at 12:28 PM on March 20, 2009 [20 favorites]


the distinction here is how often Bowman has to not only justify, but "prove" the validity of miniscule creative decisions

He gives one example in the blog post. Whether it's illustrative of a hundred other times or whether it was just one time that really stuck in his craw is speculation on my part and yours.

If you are dealing with an engineering culture that is used to seeing data trotted out, they will expect you to provide data.

I'm not sure whether it counts as irony in the strict definition, but if you read Joe Clark's blog (referred to in the post) he is constantly harping on the lack of empirical testing for design elements for things like captioning and wayfinding. Do you think Mr Clark is misguided for demanding proof on which font has the best legibility for subway platform signs? Can Mr Clark not speak a designer's language because he too demands practical proof of design claims?
posted by GuyZero at 12:34 PM on March 20, 2009 [1 favorite]


The implication on the 3 vs. 5 pixel border is that the people to whom he was justifying his decisions wanted to see data-driven proofs that one was superior to the other, and that this sort of thing was common. The problem from a design perspective is that this is an extremely reductionist way to view the details--as if a design is nothing but the sum of its parts. I think most designers would agree that the point of design is to get something greater than the sum, that a coherently designed whole cannot be justified only by justifying the details.
posted by fatbird at 12:36 PM on March 20, 2009 [4 favorites]


He gives one example in the blog post. Whether it's illustrative of a hundred other times or whether it was just one time that really stuck in his craw is speculation on my part and yours.

no it's not. He gives one example, confirms another one and says the following:

I’ve grown tired of debating such miniscule design decisions.

It's not a huge leap to assume he's grown tired of it because it's a frequent occurrence.
posted by shmegegge at 12:41 PM on March 20, 2009


or, if you prefer, here's another excerpt from the same blog entry:

Reduce each decision to a simple logic problem. Remove all subjectivity and just look at the data. Data in your favor? Ok, launch it. Data shows negative effects? Back to the drawing board. And that data eventually becomes a crutch for every decision, paralyzing the company and preventing it from making any daring design decisions.

notice key phrases like "reduce each decision... crutch for every decision... preventing it from making any daring design decisions..."
posted by shmegegge at 12:45 PM on March 20, 2009


Google (both google.com and their search appliance) has some truly awful HTML. The search appliance has, on the back end, the kind of HTML that reminds me of 1997. Tables where tables are not needed. Nested tables, sometimes for no purpose which I can discern.

Google can still be lightning-fast and also get within shouting distance of compliance. I just pulled up the HTML source for the page and selected a bit at random:

<td nowrap width=25%><font size=-2>

Aside from the fact that it should be nowrap="nowrap" as we're not supposed to do attribute minimization, they also thought "screw having quoted attributes." Oh noes, each attribute represents an entire two bytes! But then they use quotes in other spots. Run that puppy through a validator; what a mess. As long as big corporations like Google feel comfortable throwing out crap HTML with nobr tags in the mix, browsers are going to be big and bloated because they'll have to support lousy code.

I don't think that "design" necessarily means a lot of soft gradient backgrounds with a lot of frou-frou stuff and people with strange outfits and pinched expressions arguing over pixel-by-pixel placement of images, or hurling cheese at one another in heated discussions about fonts. Sometimes it just means someone who will ask for good HTML and a favicon.ico that doesn't look like the puppy stepped in little Ricky's fingerpaints again.

It's clear that Bowman's expertise was not well-used while he was there. Pity.
posted by adipocere at 12:46 PM on March 20, 2009 [4 favorites]


Personally, while I always feel bad when people say I am "MeFi's Own Jason Scott", I would advise against calling him "Mefi's Own Joe Clark". That man is toxic, toxic, toxic. Anything he sticks his face into is worse for the trauma, and his "nobody understands but meeeeeee" schtick doesn't work where it counts, that is, anywhere.

Going back on topic, there was a period where Google successfully came off as The Company Where You Wanted To Be, and of course, this was often from people who were cavalier enough in their goals and ideas to just go with The Place With Money or The Place With Cachet. But then people started leaving, which one would expect, but in the modern world of self-publishing/weblogs and a million eyeballs everywhere, these declarations of resignation are now quite public and quite high profile.

It sounds like Bowman joined Google, didn't like it after a couple years for his goal, and left. There's thousands of cases like this. I don't know if it's that big a deal.
posted by jscott at 12:47 PM on March 20, 2009 [3 favorites]


Many commenters here are mixing up different senses of the word “design.” Bowman has good coding skills but is a visual or graphic designer, and at that level, no, if you hired him as visual design lead he very fucking well does not have to justify 4 vs. 5 pixels to a nerd.
posted by joeclark at 1:06 PM on March 20, 2009 [2 favorites]


I'm not sure whether it counts as irony in the strict definition, but if you read Joe Clark's blog (referred to in the post) he is constantly harping on the lack of empirical testing for design elements for things like captioning and wayfinding. Do you think Mr Clark is misguided for demanding proof on which font has the best legibility for subway platform signs? Can Mr Clark not speak a designer's language because he too demands practical proof of design claims?

HCI is a discipline taught within Computer Science departments, not art schools. That most usability "experts" seek measurable "proofs" of the effectiveness of their work should therefore not surprise us, since most of them have absorbed a tendency to seek and provide data-driven justifications for their decisions. This is often true even when the individual designer wasn't formally trained, since by now the entire culture of usability is highly determined by those who have been formally trained. However, this does not change the fact that aesthetic questions are not susceptible to measure. Or rather, that measurement plays a different role for a properly aesthetically trained practitioner. It is never sought as some definitive determinant, but only as a subordinate aid to human judgment. (Yes, artists trained in art schools also measure - but again, from a different attitude and for different purposes than an HCI student would.) And in fact, I would argue that the constant seeking after empirical standards among the usability folks is actually a hindrance/crutch in many cases. It depends on the specific case, but in my experience, the creation of supposedly objective standards is often a stand-in for the much more difficult task of developing good aesthetic judgment.

Those with poor taste seek objective measures to prove theirs is actually provably superior to good taste.

Legibility regarding type is an ambiguously an aesthetic issue. Even so, it isn't absolute. For example, we could begin by distinguishing between typefaces that are legibile at a distance versus typefaces that are legible in fog versus typefaces that are legible for new readers of a language. By doing so, we'd begin to acknowledge that a word that we'd formerly treated as an adequate subsitute for a conceptual monolith (the absolute ideal of legibility) is a complex multiple reference. Of course, when it comes to road signage, we'd also have to begin to acknowledge that legibility is not the only concern. And thus, aesthetic issues would begin to enter more fully into our decision.

Also, I might add that engineers are fond of seizing the high ground with words like "practicality". Aesthetics is practice. And it is practical. It is real. It makes a real difference in people's lives. It's just a different practice.

It's ironic that an engineering culture that owes so much to the ancient greeks has so little of that culture's appreciation of the practical impact of good aesthetic judgment. That, in fact, "practical" is, for them, an antonym of "beautiful" - or even worse, they think practicality is a suitable and even superior stand-in for beauty. For the Greeks, this would have been an absurdity.
posted by macross city flaneur at 1:06 PM on March 20, 2009 [6 favorites]


the great web designer

On what planet? Wired has one of the most God-awful, sinfully-ugly designs on the web (at least for public, high-traffic websites). I mean, yeah it's great that you got rid of your tables for layout.

<div>But
  <div>this
    <div>doesn't
      <div>really
        <div>help
          <div>anything</div>
        </div>
      </div>
    </div>
  </div>
</div>

And hard-coded widths means anyone suffering with an 800x600 display (like some of these) has to scroll horizontally. That's a mortal sin in web design. Why couldn't the layout be more flexible, what with his incredible CSS skills?

Google can't decide on one shade of blue to use on the site, so they're experimenting with 41

Well that's just ridiculous. Clearly they should just take their edicts from some design potentate's whims, instead of experimenting. God, how crazy are those Google people?
posted by Civil_Disobedient at 1:07 PM on March 20, 2009 [2 favorites]


jscott, it is more like a question of “Who understands captioning and audio description better, you or me?” The answer is me. You keep bringing it up.

Handbags at dawn, Jason.
posted by joeclark at 1:12 PM on March 20, 2009 [1 favorite]


Macross le flâneur, we (I say “we” advisedly) tend to measure legibility strictly in terms of specific reading applications, not in general. As people like to say, blackletter type was highly legible back in its day.

To continue with that example, a trained and experienced visual designer will at least know not to compare Comic Sans and Georgia and will instead make rational comparison choices based on years of experience and knowledge which, while unquantified (that’s why we’re measuring it experimentally), exists nonetheless. This is a real example, by the way – my clipping file is chock full of investigations of plainly ridiculous typeface pairings.

This is the sort of knowledge someone like Bowman walks in the door with. And now, it seems, walks out the same door with.
posted by joeclark at 1:19 PM on March 20, 2009


Well that's just ridiculous. Clearly they should just take their edicts from some design potentate's whims, instead of experimenting. God, how crazy are those Google people?

Well, right off the bat, experimenting "with blues" is dubious. Because all color choices only begin to matter within the context of other color choices. Thus, if Google were experimenting with 41 different color pallets, that would begin to make a bit more sense. If it were further qualified by a description of the design context (ad? logo? with what specific range of applications?), it would actually become a real design question.

All this "blue experimenting" goes to show is that no one responsible for such "experiments" at Google has taken even an introductory color theory seminar or has the most basic understanding of design questions.

Of course, if somebody said that Google was experimenting with "using 1 bit instead of 2" for, uh, "things in general having to do with computer stuff" - the universal hew and cry among MeFites would go up: "What could that possibly mean?! It's such a general statement that it is laughably, utterly, nonsensical."

Yet, the idea that Google might "test their blues" goes by without a second thought. This, if anything, testifies to the abysmal state of visual aesthetic understanding among the average person.

So no, "testing out 41 blues" is not crazy. It's asinine.
posted by macross city flaneur at 1:26 PM on March 20, 2009 [1 favorite]


On what planet? Wired has one of the most God-awful, sinfully-ugly designs on the web

To be fair, Wired also has one of the most godawful, sinfully-ugly print publications as well. Whenever I see it, I'm appalled anew each time that print designers would put their name on the masthead of that abomination.
posted by Devils Rancher at 1:30 PM on March 20, 2009


All I know is Google's sites are butt ugly and therefore, obviously, they're doing something wrong.
posted by delmoi at 2:08 PM on March 20, 2009


On what planet? Wired has one of the most God-awful, sinfully-ugly designs on the web (at least for public, high-traffic websites).

What you're missing is that the 2002 redesign of wired.com was a landmark in web design. It proved that CSS-driven, standards-based design was not only possible but doable on a public-facing company's website.

Looking back now, it's clear that best practices hadn't yet developed, and it's a giant coding mess thanks to it needing to support Netscape 4/6 and IE 5/5.5. But Bowman at least proved that yes, it could be done.

That's what makes his stint at Google so disappointing. All he managed to produce of any significance was Google Calendar, which interface-wise is far more elegant than most other online calendaring systems. But when people are trying to make you intrinsically justify border widths, the equivalent of car designers arguing over whether the plastic hubcaps should have 5 or 6 spokes, it's time to say goodbye.
posted by dw at 2:08 PM on March 20, 2009 [1 favorite]


Those with poor taste seek objective measures to prove theirs is actually provably superior to good taste.

Certainly, if there's one thing everyone can agree on, it's who has good taste and who has bad taste in visual design. That there can never be a truly perfect objective measure of anything should be enough to tell us that we should never, every attempt to quantify anything, ever.

And yes, my face is as straight as yours is.
posted by GuyZero at 2:29 PM on March 20, 2009


So no, "testing out 41 blues" is not crazy. It's asinine.

Or, you know, it's wrong or misreported or being reported completely out of context.

at that level, no, if you hired him as visual design lead he very fucking well does not have to justify 4 vs. 5 pixels to a nerd

Does he have to justify it to the person who hired him and symbolically signs his paycheques? What if that person also had "good taste"? I will admit that I am projecting my own biases on Bowman's posting, but not any more than everyone else around here. Don't make the mistake of assuming that Bowman is grinding the same axe as you.

That man is toxic, toxic, toxic. Anything he sticks his face into is worse for the trauma, and his "nobody understands but meeeeeee" schtick doesn't work where it counts, that is, anywhere.

Joe, you stay crazy. And far away. Far away and crazy are a good combination.
posted by GuyZero at 2:37 PM on March 20, 2009


And that data eventually becomes a crutch for every decision, paralyzing the company and preventing it from making any daring design decisions.

As I see it, a Bowman Daring Decision is one that fares worse in the data. At which point, the person you're daring is not the company. You're daring users to stop using your design.
posted by pwnguin at 2:56 PM on March 20, 2009 [1 favorite]


Here in the UK they have whole classes of people they call "creatives". I'm working on twitter/facebook app lets you send them cockpunches. With rounded corners.
posted by srboisvert at 3:06 PM on March 20, 2009 [1 favorite]


The whole idea that once you have "The data" you can make a decision is absurd. Data itself doesn't prove anything. You have to also know that the experiments used to derive the data are themselves sound. You have to know When it comes to something as subtle as aesthetic choice then how you setup the experiments can have a huge impact on the results that you get.

And speaking of which:

A/B testing, like democracy, is the worst form of design validation except all the others that have been tried from time to time.

Do you have any actual data to back up that claim? throughout this entire thread you've been making all kinds of statements with no empirical backing at all. How do you know that "collecting data" actually produces better design results. In my experience "Design by committee" produces awful results and I can't see why anyone would expect "Design by democracy" to produce good results either. So where is the empirical evidence for your claim? Assuming Google's goal is to maximize profits, then how can you be sure that 1) The experiments that they are actually doing are well thought out and 2) That a design in some local maxima where some random subset of the population finds it least offensive is actually better then something done by an artist with a single vision?
posted by delmoi at 3:20 PM on March 20, 2009


delmoi: "the other problem is that they can easily get stuck in a local maximum this way." [5 favorites]

me, a few minutes later and failing to preview: "One problem with A/B design is that you settle onto local optima." [2 favorites]

See, A/B testing works!
posted by zippy at 3:28 PM on March 20, 2009 [4 favorites]


If I actually had a central thesis to my argument (a dubious proposition) it would be that Google does not produce art. It makes little boxes that people click on and when people click them, it makes money.

Given that thesis, art has no meaning to what Google does.

A/B testing is good for figuring out which of several designs - all of roughly equal aesthetic value for arguments sake - does a better job at making people click in a box. I wouldn't claim that A/B testing is equivalent to aesthetic judgment and if I did claim that then I'll gladly retract it.

So I don't have any evidence really. Though trying to disprove empiricism by demanding empirical evidence is a tad ironic, but whatever.

Google has limited resources and for any given tasks can produce a limited number of designs. That it produces more than one design is more than most companies do. How would you propose to decide among them, other than by fiat?
posted by GuyZero at 3:29 PM on March 20, 2009


GuyZero, design ≠ art. And I don’t know what you mean by “far away”; I’m right here in Leslieville. I just had three engineering students drop by talking about fonts and signage. Shall I make you a cup of tea?

Meanwhile, can somebody find out if Bowman really is going to Twitter?
posted by joeclark at 3:44 PM on March 20, 2009


Assuming Google's goal is to maximize profits, then how can you be sure that 1) The experiments that they are actually doing are well thought out and 2) That a design in some local maxima where some random subset of the population finds it least offensive is actually better then something done by an artist with a single vision?

Apart from working there, reading the code, analyzing the experiments and attempting to duplicate the results, I'm tempted to rely on the fact that Google hires very smart people who know to do these things. People like Douglas Bowman seem to accept the process:

I can’t fault Google for this reliance on data. And I can’t exactly point to financial failure or a shrinking number of users to prove it has done anything wrong. Billions of shareholder dollars are at stake. The company has millions of users around the world to please. That’s no easy task. Google has momentum, and its leadership found a path that works very well.
posted by pwnguin at 3:56 PM on March 20, 2009


Joe, I decamped my beloved Roncesvalles for the belly of the very beast we're discussing. And besides, Leslieville is, like, SO FAR AWAY FROM EVERYTHING! If you can't get there on the 504, it doesn't exist. You'll be going on about Oakville next or something.
posted by GuyZero at 3:57 PM on March 20, 2009


On preview, I meant to convey "apart from working there, ..., things I cannot do". I am not a Google employee, even if the previous comment can be parsed to imply that.
posted by pwnguin at 3:59 PM on March 20, 2009


On what planet? Wired has one of the most God-awful, sinfully-ugly designs on the web (at least for public, high-traffic websites).

Bowman designed the Wired site seven years ago. There have probably been coding and staffing changes in the meantime.
Seems like I knew so little back then. I was learning so fast, and was constantly discovering better methods, practices, and techniques. Immediately after the redesign, I wanted to go back and do it all over. To build it better. But I had to move on.
Wired redesign turns 3 from three 1/2 years ago.

Or, you know, it's wrong or misreported or being reported completely out of context.
A designer, Jamie Divine, had picked out a blue that everyone on his team liked. But a product manager tested a different color with users and found they were more likely to click on the toolbar if it was painted a greener shade.

As trivial as color choices might seem, clicks are a key part of Google’s revenue stream, and anything that enhances clicks means more money. Mr. Divine’s team resisted the greener hue, so Ms. Mayer split the difference by choosing a shade halfway between those of the two camps.

Her decision was diplomatic, but it also amounted to relying on her gut rather than research. Since then, she said, she has asked her team to test the 41 gradations between the competing blues to see which ones consumers might prefer.
Google has limited resources and for any given tasks can produce a limited number of designs.

All the more reason for the limit to be closer to 2 than 41. How many people are really going to be able to distinguish between those interim gradations?

That it produces more than one design is more than most companies do. How would you propose to decide among them, other than by fiat?

By splitting the difference, or by going with the greener shade that tested better.

Oh noes, each attribute represents an entire two bytes!

The rationale is that the two bytes times the millions and millions of page views adds up to huge amounts of bandwidth.
That equates to around 216 million searches per day. If they save, say, 5 bytes from their code, thats 1 080 000 000 bytes saved, which works out to around 1030 megs, which is a gig of bandwidth saved per day by not having 5 extra bytes that might make the page W3C compliant.
posted by kirkaracha at 4:01 PM on March 20, 2009


Check out the Xooglers blog for other ex-Googler stories.
posted by archagon at 4:09 PM on March 20, 2009


By splitting the difference, or by going with the greener shade that tested better.

In this instance, testing is clearly overkill and is intended to resolve a political issue and not a technical or design issue. Quite possibly this is Meyer's way to say STFU to meddlesome PMs who second-guess designers and it's not a serious test at all. In general however, testing is the only way to get a decent idea on whether people will click something more given any arbitrary change in the page layout.

I'm going to go out on a limb here and assert that most A/B tests at GOOG are more differentiated than blue versus green buttons.
posted by GuyZero at 4:11 PM on March 20, 2009


Certainly, if there's one thing everyone can agree on, it's who has good taste and who has bad taste in visual design. That there can never be a truly perfect objective measure of anything should be enough to tell us that we should never, every attempt to quantify anything, ever.

And yes, my face is as straight as yours is.



No one suggested we should "never attempt to quantify" anything. Your extreme straw man is a disappointingly sloppy response to what was a tightly circumscribed argument about the role of measurement in aesthetic decision-making and the attitude of the artist toward it. I also did not suggest the abandonment of attempts to quantify things in general, nor did I adduce the impossibility of "perfect" measurement as evidence against empiricism. In fact, it is only (naive) positivists - e.g. undergraduate science students - who think that the imprecisions and inaccuracies of specific empirical efforts are a great black mark against science. If philosophical arguments against empiricism actually depended on (accurate or inaccurate) empirical work, they would indeed be weak. No, implicit in my critique of measurement is a rather more fundamental idea - it requires no evaluation of the empirical value of measure. It is simply the assertion that we have no standard against which to judge the "perfection" of our empirical studies other than other empirical studies - and that the process of revision of one empricial study of another is precisely discursive and not "correspondent" to facts in the world. You will find even your much more rigorously trained allies struggling with this argument - and indeed, Wittgenstein, the most famous exponent of positivism, found himself capitulating to it a great deal later in his career (after he had ample time to realize the naivety of his earlier "brilliant" work). I would highly recommend you pick up some Popper, Dennet, and Feyerabend if you'd like to understand some of the contours of the debate.

However, the most important thing that your comment reveals you missed is a distinction I drew between the kind of phenomena and judgments to be made in two separate arenas. Aesthetic judgments, I implied, are qualitatively different from the ones engineers are used to making. They involve a different kind of understanding, one that isn't improved by the common resorts of empirical science. So, I rather (too) neatly gave you an out, Guy Zero. I set up a nice clear distinction between two domains of judgment, and rather simply suggested that Google keep its peanut butter out of Bowman's chocolate - or even better, stop pretending that its peanut butter is the same thing as chocolate.

Now, if you want to have an argument about epistemology and ontology, we can do that. But remember that I gave you a chance to avoid that. Also remember that snark is scant defense against reasoned argument and deeper knowledge of the subject in question.

Also, as a PPS, I would submit to you that the fact that there are disagreements about taste does not mean that some people do not have better (more informed, better developed) taste than others. Just as the fact that there are disagreements over the pace of climate change doesn't mean that there aren't more and less informed/trained opinions in ecology. My point in saying this is not to argue that aesthetic judgment and scientific knowledge are the same - it is simply to reveal the weakness of your argument about "agreement".
posted by macross city flaneur at 4:18 PM on March 20, 2009 [1 favorite]


If I actually had a central thesis to my argument (a dubious proposition) it would be that Google does not produce art. It makes little boxes that people click on and when people click them, it makes money.

Yes. Unfortunately, the fact that you have come to this place reveals the weakness of my earlier argument.

The fact is, it would nice if such category distinctions as "art" and "advertising" and "software" could save us from the real messiness of phenomena, but they can't.

However, we'd probably both agree that not having a healthy design culture at Google is not likely to hurt the company in the near term. After all, Google wasn't built on design. And every company develops its own unique culture.

However, we might consider some of the ways in which the lack of a strong design component might be shaping Google's culture so as to limit its flexibility, and thus determining its future. For instance, the slowness with which Google dipped its toe into the image-ad arena. Over time, it is this kind of inflexibility that winds up hemming a company in and speeding its path to decadence and irrelevance. A large software company that makes a conscious decision not to be well-rounded is asking for it. The comeuppance may not be swift, but that doesn't mean it isn't coming.
posted by macross city flaneur at 4:29 PM on March 20, 2009 [2 favorites]


That equates to around 216 million searches per day. If they save, say, 5 bytes from their code, thats 1 080 000 000 bytes saved, which works out to around 1030 megs, which is a gig of bandwidth saved per day by not having 5 extra bytes that might make the page W3C compliant.

Except bandwidth in bulk is cheap -- 15 cents per GB according to this page, probably lower now (and lower still for a company like Google buying it in megabulk).

$0.15 * 365.25 days = $54.79 saved. 1000 bytes removed = $10,957, or roughly 8 months of preschool daycare in the Google Kinderplex.

OTOH, if they stripped a lot of that old style code (e.g. font tags) and moved them into CSS, they could probably get more cost savings on bandwidth. Even though there'd be an initial hit from having to load the stylesheet, assuming you pull enough code out you could make it up and a bit more by having the results page rely on the same stylesheet. And the savings would increase with use of the site.

And that's as far as I'm going to go with this. Standards/accessibility discussions never end well on MeFi.
posted by dw at 4:45 PM on March 20, 2009


In general however, testing is the only way to get a decent idea on whether people will click something more given any arbitrary change in the page layout.

This is simply false. There is experience and judgment to consider.

For example, let's say the test is performed on the day the NCAA tournament begins, and one page layout has a basketball related thingie on it. Knowing the day of the test corresponds to some external event, and having the experience to realize this will impact the test is far more valuable in this instance than simply knowing the result of the test.

This is called controlling for externalities. It is what test designers do all the time.

Unfortunately, when we talk about externalities in science, we are accustomed to talking about them in the context of controlling for them, which leads us to the faulty assumption that we have (obviously) controlled for all of them.

What a designer's skill and experience means is that he understands many many minute and often (and this is key) inarticulable - externalities.

That's why his judgment is so valuable, and that's why there's a high likelihood that a well-trained expert's judgment is a far better barometer than any test.

Those who believe in the result given by the test implicitly assert that all externalities have been controlled for (even though there's no way for them to know this). While the expert knows they have not been controlled for and has specific misgivings about them - even in those instances when he cannot say what they are.

Now we come to the point where we can understand the stupidity of the positivist's demand that all externalities that count be articulated. The fact is that, when it comes to many aesthetic judgments, the number of relevant externalities, the complexity of their structure, or their sensitivity to changing conditions, is simply too great for the designer to say them, write them, or test them. People who work in this field KNOW this. It's why they don't expect anyone to waste the time in testing them. Positivists, on the other hand, are too stupid to realize just how much complexity lurks out of their simple-minded purview, so for them tests are God. But this is nothing more than provincial hubris. It's idiocy masquerading as rigor.
posted by macross city flaneur at 4:53 PM on March 20, 2009 [1 favorite]


Blah, too much to reply to point by point.

a) You don't have to talk down to me, I have no pretensions to arguing epistemology and ontology. If you want to show off that I brought a knife to a gun fight, you win. Pick up your prize on the way out.

b) I'm not rejecting aesthetic judgment. But given that many, many people lack it, testing is generally more reliable - especially when it's something trivially quantifiable like generating clicks. But by no means am I saying it is a panacea or the sole way to approach things. Google obviously needs people who understand design but when it comes to generating maximum clicks it's not enough. I doubt you'd disagree with the idea that both are necessary. I am an advocate of testing because I have seen too many people who overestimate their own design ability when it comes to making pages that generate clicks. But it's not a dichotomy.

c) Certainly experiment design is a skill and it's probably hard to quantify. A bad experiment is probably worse than bad aesthetic judgment. But let's assume the best that the people at Google are competent and can design a meaningful experiment. You seem to be supposing that positivists are going to do only dumb stuff while all people going on aesthetic judgment are very, very good. It's a bit of a straw man to argue against bad experiments.

the stupidity of the positivist's demand that all externalities that count be articulated

I think you're having an argument with someone who isn't here.
posted by GuyZero at 5:21 PM on March 20, 2009


if they stripped a lot of that old style code (e.g. font tags) and moved them into CSS, they could probably get more cost savings on bandwidth

They use CSS-a-plenty. And I didn't notice any FONT tags when I last looked at the Gmail client, but I do remember a veritable rat's nest of embedded DIVs and nested TABLEs. But their stuff is consistent… try it on a slew of different platforms and it looks basically the same. That's not easy.

The designs are simple, functional, and not terribly pretty. The UI, on the other hand, is awesome. Google's web designers pull out all the stops to get the interface to behave like a "regular" application. The Gmail inbox is a perfect example. Click one row, hold down the shift key and click 10 rows down… works just like a stand-alone app. All their stuff shows this kinds of attention to detail. And you should open FireBug sometime and watch all the XMLHttpRequests it shoots off. Same with Google Maps. Just an amazingly powerful interface.

The rationale is that the two bytes times the millions and millions of page views adds up to huge amounts of bandwidth.

With the insane volume of asynchronous traffic their apps produce, it honestly doesn't look like they care too much about saving a couple bytes of traffic.

How many people are really going to be able to distinguish between those interim gradations?

Yeah, but how many companies even bother to run usability studies before releasing software or websites? You can probably count them on a couple of hands. OSS projects, maybe one hand. Maybe. If Google's got the metric assload of resources to run psychological studies on 41 shades of blue, well… great! Wouldn't you rather they do that then, I dunno, buy a new helicopter for their execs?
posted by Civil_Disobedient at 5:21 PM on March 20, 2009


You seem to be supposing that positivists are going to do only dumb stuff while all people going on aesthetic judgment are very, very good. It's a bit of a straw man to argue against bad experiments.

Whether they are positivists or not, people who fail to recognize the value of individual human skill and experience over any testing process are fools. Implicit in the idea that testing procedures are in any way competitive with human judgment is the assumption that all relevant externalities have been controlled for or can be controlled for. In other words, every experiment is a "bad" experiment in the sense that it only ever tests for linear, articulable variables - while human brains and sensory systems are designed to deal with massive parallel data in the real world, in real time.

Manager-engineers who value process over people may have a democratic, populist wish that all people are the same, that all judgment is the same, etc. They may even want to force it to be that way on their watch. But wishful thinking is the source of much error. Anyone who told you they knew they could translate their design skills into measurable clicks was also wishful - but you were equally naive for having so narrow a goal, if that was in fact your goal.

No one I have ever worked with in the real world has had such a single-minded and narrow idea about what design could do for them.
posted by macross city flaneur at 5:40 PM on March 20, 2009


I really like Google's visual aesthetic. It's unobtrusive, which is what I want from a set of tools.
posted by brundlefly at 6:11 PM on March 20, 2009


[comment removed - do not do that here - you have metatalk, use it, thanks]
posted by jessamyn at 7:08 PM on March 20, 2009


So I don't have any evidence really. Though trying to disprove empiricism by demanding empirical evidence is a tad ironic, but whatever.

First of all, it's much less Ironic then demanding that "empiricism" works without any empirical data. And second of all, simply demanding that you have "data" to back up your claims does not actually result in good claims. You have to know that you have the right data and that you're asking and answering the right questions. That's something that a lot of people fail to understand, especially (I would imagine) outside of academia where you don't have peer review or people who really understand experimental design doing "experiments" and then looking for correlations or whatever.

One example would be political polling where the results can be hugely skewed depending on how you ask questions, half the people dealing with the data don't even know how margins of error work, etc. Just look at Mark Penn in the Hillary Clinton campaign, who would win arguments by saying "the data" backed up his claims while losing the campaign.

Bad data can be as powerful a weapon as bad arguments, and just as easy to come up with.
posted by delmoi at 9:30 PM on March 20, 2009 [2 favorites]


"41 gradations between the competing blues"

That's my favorite Ornette Coleman tune.
posted by krinklyfig at 10:45 PM on March 20, 2009 [1 favorite]


"which is a gig of bandwidth saved per day by not having 5 extra bytes that might make the page W3C compliant."

Google can't afford a gig of bandwidth a day? Are they on the Verizon cell plan?

I understand, you know, corporate decision making. Tiny details amount to large differences in aggregate, and those numbers impress investors. But it's also what makes it frustrating. Doing it right succumbs to doing it cheap. But the another view is that doing it right is going to be more beneficial in the long run. It's very difficult to fully embrace that philosophy when a corporation gets big enough, especially if it's public.
posted by krinklyfig at 10:54 PM on March 20, 2009


The discussion between Macross City Flaneur and GuyZero is fascinating, as it basically captures a debate between two schools of thought that one often sees at HCI conferences. The first school of thought might be characterized with a famous software engineering quote: "You can't test quality into a software project." The other school of thought might be characterized as: "Without testing, how do you know you have quality?"

As someone who teaches and does research in HCI, these two perspectives are ones we run into daily. There are simply too many variables to account for in any given design, but unless we can account for the most important ones, how do we know if our design is effective in achieving its goals? Or as Feynman might say, how do we know we're not fooling ourselves? I think there are deep truths to both of these perspectives, and the challenge here is knowing when to apply one versus the other, as well as how to combine these two views to create an effective design.

Kicking up the discussion a level, I think it's also worth talking about what kinds of new opportunities large scale data sets offers for design and user interface engineering. I saw this amazing talk last February about what Google can do with large sets of data from virtual testing. Using Google Maps as an example, the presenter showed relatively small changes to the user interface on each progressive slide. Should the search results be shown on the left side or right side of the map? Should pictures of the street location be included or not? Should it be called Google Local Search or Google Maps?

While all of these were what we might consider micro-optimizations, the results were quite compelling, in that Google could show a sizeable increase in clickthru rates for a given design, even for things that might seem like trivial design decisions. In a room full of HCI experts, the presenter showed a screenshot of another page, and asked which button label would lead to the best results. Most of us guessed the 2nd or 3rd best label, which turned out to have at least 5% fewer clickthrus than the best one.

I would agree that there is a very strong risk of settling into a local optimum using this approach alone. Realistically, it's better to view this kind of virtual testing and large-scale data sets as another powerful new tool among a set of tools. As such, while I strongly sympathize with Bowman's view of not wanting to be locked into a process that relies solely on empirical data, I would also consider it foolish not to make use of such a powerful technique that can give us new insights into the problems we (both designers and engineers) are trying to solve.
posted by jasonhong at 11:00 PM on March 20, 2009 [3 favorites]


Minor clarification: I don't know if Bowman did or didn't use virtual testing and large sets of data, but generally, I would consider any user interface designer foolish if they didn't make use of this kind of technique if it were easily available to them.
posted by jasonhong at 11:03 PM on March 20, 2009


I've long wondered if Google was an experiment in function over form, in some ham-handed attempt to "prove" to designers everywhere "see? people don't give a shit how ugly things are. Look at our logo for fuck's sake. A 6-year old could make that in an afternoon using a Mac LC and Superpaint!".

My arguments toward good design aesthetics, and I'm definitely siding with Bowman here as I've been in his position many times before, is thus :

1. Why does some higher-up's opinion, with no formal training or knowledge in the field of design, hold more weight than a given designer's years of experience, (dare I say "expertise")? If the answer is "cause that person writes the paychecks" - that's someone pulling a power trip and it's a bit unfair to point the whimpering-prima-donna-finger at the designer in that scenario. However, it happens all the time.

2. I will put up no argument that some people, maybe even most people, care very little about how a thing looks so long as it functions well. Hey, that's great. But, why appeal to the lowest common denominator? In this case, I think most people agree that Google does what Google does very well and makes some very solid products. If they beefed up the design side of it - those who currently use it will continue to do so (cause, hey, they don't care, right?) and maybe... just maybe... they'll put a smile on the face of the more discerning crowd as well. Win-Win, no? Moreover, why is striving for a pleasant visual experience something that's being looked down upon?

3. Design isn't purely subjective, and it's fairly easy to "prove" certain points. For instance, this site is much easier on the eyes than, say, this site. When I say "easier on the eyes", I don't mean "pretty". I mean.. sit in a room and stare at Apple.com for a minute or two solid and then repeat that process with the 2nd link and tell me which one actually leaves behind a short burn-in effect on your retina when you look over to a wall (or another website, or another part of your screen). Which one feels more inviting? Makes you feel like a company really put their best foot forward to show they care? That's where good design comes in. These things matter.

I don't know that Twitter having a better design lead matters nearly as much as Google having lost one. Far as I know, most people using Twitter don't even view the site itself, but merely use any of 1000+ apps that post to Twitter through an interface they prefer. That's not Bowman's concern, of course, so kudos to him for finding a better fit.
posted by revmitcz at 11:23 PM on March 20, 2009


I would agree that there is a very strong risk of settling into a local optimum using this approach alone. Realistically, it's better to view this kind of virtual testing and large-scale data sets as another powerful new tool among a set of tools. As such, while I strongly sympathize with Bowman's view of not wanting to be locked into a process that relies solely on empirical data, I would also consider it foolish not to make use of such a powerful technique that can give us new insights into the problems we (both designers and engineers) are trying to solve.

I don't know if Bowman did or didn't use virtual testing and large sets of data, but generally, I would consider any user interface designer foolish if they didn't make use of this kind of technique if it were easily available to them.


By and large, I would agree that the problem is not the data per se as much as it is the attitude towards the data. The caveat to this being that we need to increase our level of self-consciousness about the epistemological and psychological tendencies that are set in motion by the existence of the data itself.

The comparison I would make is to presidential polling. There are many who think that on balance the very existence of polls is negative. Especially in the days after they were first coming to influence, the temptation to react strongly to the polls and to treat them as gospel was very great. Immature use of polls really colored the Clinton presidency - a reactive, attention-deficit presidency. Bush II represented another kind of reaction - driven by the realization that selecting which polls to present and always insuring that there was a pollster who had data that favored the president's position was key to making the polling process "work". In other words, the realization that the polling process was itself a highly manipulable thing - essentially numerical rhetoric - that was effective precisely because it gave the plausible appearance of objectivity. We'll have to wait to see exactly what role polls play in the Obama presidency, but rest assured that we are just beginning to mature in our use of polling data.

But the arc or "story" of data is bound to be the same in the enterprise as it is in the media-sphere. Gradually people realize there is no more objectivity in numbers than there is in the shrieks of guests on Jerry Springer. Today, "you have your opinion and I have mine" is a cliche. Tomorrow's cliche is "you have your data and I have mine".

It is only the relative plausibility of the "objectivity" of data that makes it a "powerful" tool.

For example, the concept of "clickthrough" is a free-floating, decontextualized fantasy. What does it really mean? Absolutely nothing. In order to analyze "clickthrough", we'd really have to know more about the specific case. This is another of those monolithic concepts - supposedly always a "good" thing (yay! clicking!) - that immediately disintegrates under closer scrutiny. People might be clicking because they are hoping to see information they should be getting from the screen they're already looking at. They might be clicking because they're frustrated. There is no "objective", unambiguous meaning of clickthrough - though again, the ignorant will always assume there is one if it helps their cause, and then call it "objective".

Context is king. Human experience, savvy, and judgment is king. Asking the right questions is king. Data is a pawn.

Here is a summary of 100 books/news stories written last year, as it will be of 100 next year. "It is often assumed that the scientific evidence in support of X is definitive, but the reality is more complicated. Speaking with Professor Z of University Y about his field, I was reminded that the conclusions of working scholars is often far more qualified, provisional, and tentative than the popular literature makes it appear. Upon closer examination, it would appear that even sub-thesis N of X, one of the most supposedly rock solid ideas in the entire field, is the subject of great debate among scholars."

What data is good for is mostly convincing people to do what you've already decided is for the best. There is no data that speaks unambiguously to any particular issue. It must be framed, interpreted, represented. Every study has flaws, and the enemies of the conclusions a given study was designed to support will always point them out, while the proponents of those conclusions will minimize them.

Just ask Karl Rove.
posted by macross city flaneur at 12:18 AM on March 21, 2009


Now that I can participate in this forum without being flayed alive, let me expand on GuyZero at 3:34 and say I am a big fan of testing of graphic design. (I’ve written a lot about that on my blog.) But as described by Bowman, they’re testing the wrong things and just not trusting him on what should and should not be tested. They did hire him as visual-design lead; they didn’t allow him to lead.
posted by joeclark at 5:01 AM on March 21, 2009 [1 favorite]


Personally, while I always feel bad when people say I am "MeFi's Own Jason Scott", I would advise against calling him "Mefi's Own Joe Clark".

Metafilter: Don't call him "Mefi's Own".
posted by the cydonian at 6:16 AM on March 21, 2009


let me expand on GuyZero at 3:34 and say

The timestamps are adjusted for the user's timezone.
posted by delmoi at 11:34 PM on March 21, 2009


What a great idea, to leave google. Everyone knows that the next big search engine is gonna come out six months from now, and google will be deserted. It's the social networking sites that will stick around for years and years and become set pieces of the internet.
posted by tehloki at 1:28 AM on March 22, 2009


Design is increasingly being treated as a pseudo-science instead of what it primarily is: art.

Designers need to stop jumping onto pseudo-scientific ideas, borrowing borderline meaningless concepts and fancy words from philosophy (this thread is full of them), or pretending there is any real form of objectivity in the subject that is used justify their choices. In reality, design owes more to trends than anything else and I've noticed that recently, particularly on the web, people attempt to deny it.

Engineers need to stop trying to inappropriately treat design as a rigorous science where everything can be objectively measured, and to understand what can and can't be done, and leave it to the designers to make the art decisions.
posted by HaloMan at 8:49 AM on March 22, 2009


I want whoever changed the GoogleMaps UI fired. It used to be so much better a year or two ago.
posted by Eideteker at 12:29 PM on March 22, 2009


Design is increasingly being treated as a pseudo-science instead of what it primarily is: art.

Except with the web it isn't purely art. The science comes into play because that art often requires human interaction. And that means web design isn't ars gratia artis but "form follows function." It's architecture, not painting.

And every good architect knows how buildings function, how they're made, the advantages and disadvantages of the materials they're made of, and the physics that keeps the building from crumbling. In the case of the web, it's about understanding the materials (XHTML, CSS, JavaScript, Flash) and how people interact with the materials (user experience, user-centered design, expected behavior, etc.) So it is science as much as it is art.

That does not excuse the ludicrous notion that you have to scientfically test how wide a border should be. There is nothing in any architecture book that states the best practice for putting up house numbers. That's a design decision. Just like border width.
posted by dw at 3:19 PM on March 22, 2009 [1 favorite]


Engineers are asked to prove their case for adding a single byte of data to certain data structures at Google. To VPs. Yet Bowman should be allowed to do whatever he wants, anytime? Everybody is accountable for their work

Erm, this is not my experience. Of course, like any big company, different groups work differently. And my part of Google (YouTube) is a little more isolated -- we certainly don't work the way you describe, only for large design (here I'm talking code design/architecture, not visual design) changes would I feel the need to seek out detailed review (and I could check such a change in to the live branch without it -- but no matter how good you are, it's best to get some other opinions before you do a major change). Engineers I know on other projects have varying degrees of this.

There are known exceptions -- famously, the home page is controlled down to the byte, pretty much. And I'm sure Bowman did have the experiences he describes. But the idea that everything at Google works this way just doesn't fit what I see. Many of the engineers I know on other projects that also don't have this hyper-controlled environment are unreleased projects, and that is also a factor (the risk of a change is much lower). For something as important as Search, even a small change could be very bad, so it makes more sense to treat it differently.

But it's true, experimenting and data collecting is part of the core culture here, and that's not going to change. However, not all groups and people take it to the extremes presented in some of these anecdotes.
posted by wildcrdj at 7:37 PM on March 22, 2009


Someone linked this in another post, but not here:

Dear Blogosphere: Resist the urge to take what few ‘inside baseball’ tidbits come from folks departing Google (or anywhere else, for that matter) and extrapolate the interior nature of the company from those small gleanings. On a numbers basis we don’t comprise statistical significance and our choice to depart is clearly a confounding factor. You’re trying to judge a group by their least satisfied (and most vocal) 5%. The data you need just isn’t there.
posted by GuyZero at 5:14 PM on March 24, 2009


Twitterization confirmed.
posted by joeclark at 2:43 PM on March 31, 2009


« Older Bats Flying in Slow Motion   |   Equine achondroplastic... Newer »


This thread has been archived and is closed to new comments



Post