Siri’s great at some things (fetching the weather), reasonably good at others (setting up reminders), and relatively hopeless at many other things (its fallback to anything it doesn’t know how to do is to suggest a web search). I speak fluidly in an unremarkable Midwest American accent, but I still have to take care to enunciate.

There’s a large group of people for whom Siri is almost completely useless, however: many deaf people, and many people with speech impediments. It’s also far less reliable for English speakers with certain accents.

Estimating the number of deaf Americans is a little complicated. People can develop hearing loss at different points in life, and estimates of people who speak American Sign Language vary wildly, from 100,000 to 15 million. There are more reliable figures on people with speech impediments: roughly 5% of people at some point in their lives, and about 1% of Americans at any time.

The Siri problem strikes close to home for me, since members of my immediate family stutter and find voice interfaces (like call center menus) unworkable. They were offended that, while the iPhone offers a keyboard interface just about everywhere else, you can’t pull up a keyboard for Siri. It’s not available as an accessibility option under “Settings” either.

You can type to correct an entry if Siri misunderstands you, but this functionality is hidden. The spoken text appears in the same font and background as the non-editable computer responses. There’s no visual cue that you can touch your request to bring up a keyboard. I only discovered this functionality weeks later while researching Siri for this post.

Why is accessibility important? Isn’t Siri “good enough” if it works for 90%-95% of Americans?

I believe fundamentally that technology should open doors to people. Siri has huge potential as accessibility software, making the iPhone’s functions easier to use for those with motor or visual difficulties. But an accessibility gain for one group shouldn’t mean an accessibility loss for another. That’s moving in the wrong direction.

This is eloquently expressed by Dr. Don G of deafhooddiscourses.com, who notes that technology like cash registers and films that started off as accessible to deaf people later became inaccessible. (Note to hearing users: click the closed captioning (“CC”) button at the bottom right of the video for captioning in English.)

Dr. Don G otherwise really likes his iPhone 4S, as do I. For what it’s worth, Siri was released as “beta” (an unusual step for Apple), so I have high hopes that they’ll add full keyboard support in an update. It’s the right thing to do.

If you’re interested in government accessibility guidelines (a.k.a. Section 508), including providing alternative input methods to speech (§ 1194.31(e)), they can be found here.

As a reminder: Loughner is accused of killing 6 people in Tucson in January and injuring 13 others, including a member of Congress. He’s schizophrenic, and court battles have raged over whether or not he should be forcibly medicated.

If his mental illness is controlled, he could be competent to stand trial and face the death penalty.

Here’s the chart that accompanied the original article:

I often say that data should tell a story, and this chart certainly does:

• At the start of the timeline, Loughner was agitated (orange) or anxious (green) roughly 70% of the time.
• His medication was resumed on July 18th (the white bar).
• After that, his depression (red) soared, to where he is depressed around 90% of the time, and numb (“bland”) the rest.

Considering the severity of the charges, severe depression seems the most sane response, and evidence that the anti-psychotic drugs are working.

I’ve never seen a chart like this, and it brought numerous questions to mind:

• What are the definitions of moods? How does “calm” differ from “bland”, and “agitated” from “anxious”?
• Did they record his mood at regular intervals, or estimate percentages at the end of the day?
• What were the trends prior to July 13, when he was not medicated?
• What privacy rights do institutionalized people accused of crimes have regarding their psychological data?

Of course, the biggest unanswered question is this:

• Does Loughner’s depression spring from remorse, or just concern for his own fate?

To be honest, I felt a little guilty seeing someone’s psychological state laid bare. I have to wonder what this chart would look like for others with mental illness, or for psychologically healthy people. What would it look like for me?

Here’s an example of what I mean. This chart appeared in the Wall Street Journal a few weeks ago, in an article about Google’s acquisition of Motorola:

Is it a solar system model? A railroad map?

Here are all the types of data this graphic is trying to show at once:

• Market capitalization
• Specific partnerships in the smartphone industry
• General levels of corporate competition with Google
• Market share of smartphones for April-June 2011
• Market share of mobile OS for April-June 2011
• Market share of set-top boxes for January-December 2010

That’s six separate things, which are being conveyed through bubble sizes, colors, connecting lines, icons, dollar figures, and percentages. Some are in different units, or for different time periods. Much of the data aren’t relevant for certain players (for instance, Cisco doesn’t make smartphones).

The first step in sorting out this mess is to put the data into a table. In some cases, a table may be all you need.

The first thing that becomes apparent is that there is some extraneous information. The chart is ostensibly about the “smartphone universe”, but HP and Cisco have virtually no smartphone or mobile OS market share. Also, what do “set-top boxes” have to do with smartphones, especially since almost 80% of the market is handled by “Others”?

The chart can be pared down like this:

The article is about the Google-Motorola alliance. So we can add some subtotals to the table:

Grouping the data provides some ideas for graphs. We’re looking at three different metrics (market capitalization, smartphone market share, and mobile OS market share). A panel chart would allow us to look at the three different metrics side by side, without trying to squeeze them all into a single complex visualization.

From this panel chart, these stories emerge:

• There are three big players in terms of combined market capitalization. Does RIM have the resources to compete?
• Google and its allies lead the pack in smartphone market share, but Microsoft’s deal with Nokia makes them a close second. There are many other small players not shown.
• Google is far ahead in terms of mobile OS market share. Microsoft, Apple, and RIM are fighting to gain ground.

The one thing the chart doesn’t make clear is that the Microsoft/Nokia strength in mobile OS market share is due to Nokia’s Symbian OS, which it is retiring in favor of Windows Phone 7.

The most important thing when designing a visualization is to keep it simple. Bar charts are great for showing one metric; scatterplots can look at the relationship between two metrics. You need to be very careful when trying to show three or more things at once; often, you’ll wind up with a confused mess. When in doubt, show less.

For instance, you could have a character walk up to his friends in a coffee shop, say “I just saw a bank get robbed,” and then describe how the robbery went down. But it’s much more exciting to go back in time and show the bank robbery in progress.

The same thing holds true in the computer world. It’s much more meaningful to show your problem to the help desk than to try to explain it to them. Anyone who has gone back and forth over email with the help desk (or tried to describe a problem over the phone) knows what I mean.

I was really impressed with the way Google+ handles its “Send Feedback” link. It’s easy to find on the bottom-right corner of the screen, and it doesn’t move when you scroll. When you click it, Google creates an interactive snapshot of the screen and allows you to highlight the problem and black out anything irrelevant or personal.

When you finish marking the page up, you can preview your work and your comments:

It’s not quite perfect. My biggest gripe: there’s no way to get from Preview back to the markup page if you want to change something. The only two options are to submit it or cancel (losing all your work).

Still, it’s a heck of a lot more elegant and intuitive than Facebook’s feedback page. This monstrosity is hidden several clicks away, is two pages long, requires you to pick a category from a lengthy list, and mashes up the feedback submission with searching for help and changing your language.

There’s been a lot of talk about what Google+ brings to the table that Facebook doesn’t already provide. So far, the most impressive thing for me (in addition to the intuitive beauty of Circles) is the clean, sparse design that contrasts starkly with Facebook’s cluttered screens. Google+ has a long way to go before it topples Facebook, but putting a great feedback system front and center is a good start.

Thanks to my coworkers David Burke and Patrik Outericky for the concept for this post.

We laughed, too, but the expression stuck in our mind. After that, we found ourselves wondering what camels and goats actually cost.

That’s why I got a kick out of the website for Heifer International. The charity works to address world hunger by giving communities livestock and tree seedlings. Here’s the ad for it starring food nerd Alton Brown:

The clever part is that the donation page is organized by animal, with donation amounts going to buy cows, sheep, pigs and, yes, camels and goats. Per the site, a camel is actually the most expensive animal you can give: $850. A goat is $120, meaning you can buy 7 goats for the price of a camel.

Digging into the FAQ, I learned that the money doesn’t literally buy an animal at that price. It really goes into funds pegged for certain types of livestock. But I prefer the idea of clicking the payment button and literally sending a llama.

And then there are numbers that just exist because, hey, people love to count.

If you missed it, New York Yankees player Derek Jeter notched his 3,000th hit on Saturday in a game against the Tampa Bay Rays. He’s not the first guy to do it; he’s actually the 28th. But it’s still a nice, big, round number that no one on the New York Yankees had ever reached, and it’s an accomplishment that happens infrequently enough that it sets sports fans’ hearts a-flutter.

It’s a number with enough resonance that it deserves to be emblazoned on a t-shirt with no explanation of its meaning.

The repetition of the number in recent weeks brought to mind a 2004 comedy called Mr. 3000. The plot involves a selfish baseball player (Bernie Mac) who retires when he reaches the 3,000 milestone… only to learn years later that, because of a clerical error, he only had 2,997 hits. It’s funny to think someone might tie their entire sense of self-worth to a number–though, to read Jeter’s interviews after the fact, it certainly sounds like he’d been fretting over it, too.

In our lives, we rarely have chances to hit the big numbers. Seeing your car’s odometer hit 100,000 miles (as mine did last fall) is that sort of event. No one will put it on a t-shirt, but it’s still worthy of a small pat on the back.

Sometimes I find myself dwelling on all the mundane things we don’t count, and wishing we somehow could. Wouldn’t it be fun to know the 3,000th time you kissed your wife?

Or maybe it would be dispiriting. If you don’t like your job, it be depressing to recognize the 3,000th morning you walked into your office building.

A few years ago, we were exhorted to walk 10,000 steps a day to stay in shape. The number itself didn’t have a lot of scientific backup, but was big and round. Yet it never seemed to get much traction because, well, it’s not much of a milestone if you have to do it every day.

As we track more and more of our personal data, there may be more opportunities to recognize personal milestones in numbers. An exercise app might alert you to the 1,000th mile you’ve run, for instance.

So I suppose I should amend what I said earlier. These numbers do have meaning: a keep-it-up message. Keep running. Keep kissing your wife. Keep driving that old car until it croaks. For Jeter, keep hitting that baseball.

What milestones would you be interested in observing?

So what are the results of using those BI tools? Well, by utilizing the site, and MPG gauge, my average MPG’s since the post have been 37.9; over 4 mpg higher. Which equates to a savings of about 40 miles per tank, or one gallon of gas (as of this writing, that’s about $3.60). Now – because the car is pretty efficient anyway, $3.60 isn’t a huge deal, but I do fill up about 2x per week (long commutes). So it’s about $7.20 – $8 per week. Which is about a 10% saving in fuel costs.

I’ve also started tracking gas stations I purchase gas from – and while that data set is still small and this inference is probably premature – but Sunoco gas yields the worst MPG’s. While my combined average for Hess/Shell/Mobil (Shell being the clear leader) is about 38, Sunoco is about 34. More to come on that.

What it really boils down to is 3 things:

1. Used correctly, BI helps reduce costs (even in everyday life) – 10% in my case or about $400 annually

2. Charts and graphs alone aren’t enough.  Understanding which metrics to review (real time MPG, supporting data – which gas stations, driving style) is much more important.

3. And probably most important (and definitely the most difficult), is knowing how to act on the insights BI provides (buying gas at the right place, driving efficiently). It’s all well and good to collect, report and look at numbers/charts/pivots, but until you turn them into actions, they don’t provide much value (not changing anything is sometimes the right action too!)

For reference purposes, this is me:

I always complain that personal data is scattered willy-nilly across the Internet. So why is Meetup clueless about my gender and ethnicity?

The short answer is, they never asked, and you can’t tell them even if you want to. Under Account, you can only specify location, language, and hometown.

Instead of asking for demographics, Meetup has you tag your interests. At some point, I told them I was interested in groups about “New in Town”, “Wine”, and “Coffee”:

These are the tags on the Asian Girls/Ladies group:

I evidently got targeted because I also have the “New In Town” tag. Meetup confirmed this over Twitter:

The tags give group owners a way to broaden their reach, but aren’t that useful for narrowing it. No one is specifically asked their ethnicity, so you can’t depend on people who are Asian to have the “Asian” tag in their profile.

The tagging system works great for generic interests (wine, hiking, knitting, etc.) where demographics shouldn’t matter. But those aren’t the only kinds of groups people are forming. In fact, this is the list of “Other Nearby Meetups” on the same page:

With the exception of ZUMBA, all of the recommended groups have a demographic component:

• Ethnic (Chinese, Men of Color)
• Religious (Jewish)
• Marital Status (Singles)
• Gender (Women, particularly women with children)

Refreshing the page brings up new selections, but these are consistently a mix of demographic-specific and demographic-blind groups.

Meetup needs to understand their users and their needs, and get the data to serve them. Many users  want a simple way to self-identify so that the right groups are targeted to them. Moreover, they want to avoid being accidentally spammed  with email inviting them to join irrelevant groups. The essence of search (and advertising, for that matter) is presenting people with relevant messages.

It’s hard, on a long-established site, to convince users to update their profiles and give you more information. But there’s no substitute for it, if that data is important. You need to ask the right questions to deliver value to users.

I certainly never thought of them as a way to rate and compare rice cookers. But at a home goods retail store in the Akihabara district of Tokyo, that’s exactly how they’re used.

Radar Charts for Rice Cookers. Sorry for the glare, I was taking a picture of a price tag in a retail store... generally not looked upon kindly in any culture.

Each rice cooker was rated from 1 to 5 on six different features.  Those ratings are then averaged up to a score for that rice cooker.

At least I think that’s what the radar charts are for.  I speak tragically little Japanese and can only read the word for yen.
This little guy pictured here (Mitusbishi NJ-18GE7-S) appears to be rated 3.7 overall with two features receiving a 5.  At nearly $450, it better be good (though it appears to be on sale until June 3rd for $250).

Perhaps of interest to chart rule followers is that the radar chart has both vertical and horizontal labels. Before all of you Stephen Few drones get worked up into a tizzy, remember that Japanese text is written both horizontally or vertically.  In fact, when it’s written vertically in books, the books are often read right to left (both the columns of text on the page and the pages in the book).

The radar chart would seem to be a good way to visually show which features are particular cooker by using the yellow highlighted section from to show where the manufacturer made price-feature trade-offs.  Although that really only works if the adjacent features are related since the radar chart creates an area of yellow between each pair of features.  If “taste” and “build quality” are two adjacent features, I’m really not sure what the yellow highlight would represent.

I guess now I’ll have to file radar charts under, “visualizations as anime.”

For the near future, most people are going to shop for a plain-old gasoline powered car. So here’s an example of the new label from the EPA’s website, fueleconomy.gov:

Here’s what I like about this label:

• The numbers are big. The three biggest numbers, which users’ eyes are drawn to, are MPG (a very familiar figure), annual fuel cost, and savings. The dollar amounts are crucial for making decisions. Note that if you shop for a gas guzzler, the top right figure reads, “You spend $xxx more in fuel costs over 5 years compared to the average new vehicle.”
• It includes a “gallons per 100 miles” figure. I advocated this in an earlier post, because you get much greater savings from small MPG increases in large, inefficient vehicles than in smaller cars.
• The source for the dollar estimates is written at the bottom: “based on 15,000 miles per year at $3.70 per gallon”. No fuzzy math.
• Next to the MPG rating, the sticker lists a range for the vehicle class (“Small SUVs range from 16 to 32 MPG”). This makes it easier to compare vehicles, though a visual scale would also be good.
• The fuel economy and smog ratings are on 1-to-10 scales that are easy to understand. The logic behind the ratings isn’t printed but is easy to find on the site. These scales replaced the government’s plan to issue letter grades (A to D), which proved too controversial.
• The website is in big letters. It’s easy to navigate, and it lets you do side-by-side comparisons between vehicles. Here’s one between a Toyota Prius and a Toyota Corolla:


The information continues for another screen, comparing the cars’ safety ratings, transmissions, passenger and luggage space, and so on.

• Finally, the QR code is a nice touch. This one shoots you to the EPA site, formatted for mobile devices. Few people use QR codes today, but with almost everyone carrying a smartphone, they’ll become more and more popular.

Overall, it’s a very sharp design that gets the right numbers in front of car shoppers. We’ll probably be replacing our 2001 Toyota Echo (a.k.a. “the clown car”) in the next few years, and I’m glad to have these new tools.