Thursday, June 30, 2011

Find QR Code Mistakes Before Making Your Own

Hat embroidered with a QR code.

The local AAF chapter here in Buffalo, Advertising Club of Buffalo, presented a primer on social media in its monthly AdLab event a couple nights ago. At the request of members, QR codes made an appearance on the topic list at the end of the presentation. While I could discuss the curious association QR codes seem to have with social media, that's outside the scope of what I want to discuss.

Technology issues aside, the presenters tried to explain to attendees that QR codes themselves are a call to action, and just as any call to action you have to be prepared to support it. One audience member suggested that just sending someone to a web site was a bad idea, another tweeted that they should be ignored:

Bottom line: forget about QR codes #adlab

In general, I think the crowd missed the point. And I think that's ok. If they don't immediately see the value of a QR code then they probably haven't identified a business case where it applies. I haven't identified a business case for a blimp for my business, so I can pretty much forget about blimps. I think we can all agree that's a safe way for me to go.

I have written about uses for QR codes before, but after my experience on Tuesday I am taking a different approach and writing up some practical, real-life experiences of where they fail in the hopes that I can provide some cautions when a real business case for their use does present itself.

Most people who will use a QR code will be using a smartphone. You don't see too many people wrangle their netbook or laptop out of the case just to capture a QR code in a print ad at their local coffee shop. Some might even have their tablet computer (ok, iPad) handy and will use that instead (ok, iPad 2).

Mobile Unfriendly

This ad promotes a series of local events that spanned a few days over assorted venues. As a smartphone user, I was looking forward to the prospect of maps, timelines, and descriptions of all the events, though I really only expected the same web site I could view on my desktop computer.

Image of a newspaper ad with a QR code.

But this is what I saw on my phone:

The web site loaded from the QR code.

And this what I saw on an iPad (I am showing the contact page to demonstrate that it is definitely not the correct site):

For context, here is the site as it appears in my Flash-enabled desktop computer (note, no handshake stock image):

Screen capture of

I reached out to the organization to tell them there were problems, and they were quick to address as best they could, but they were also limited by the platform they chose. They used a platform called Wix, which allows clients to make Flash-based web sites for free, and automatically shunts users to an HTML-powered mobile template if it senses the user is on a smartphone. The problem is that for non-technical clients, it is not clear that there is a mobile site nor is it clear how to maintain it. This is a classic example of You Get What You Pay For.

Not Mobile Optimized

The same newspaper where I found the QR code I discuss above had another advertisement for a local college promoting its open house. The text below the QR code directs readers to scan the code to RSVP for the open house.

Advertisement using a QR code.

This is what I saw on my smartphone:

The web site loaded from the QR code.

Here's what they did right with this: in the absence of being able to make a mobile-friendly site, they did nothing at all. It's not a Flash-based site, and it's not so reliant on cutting edge technology that a smartphone cannot render it. In fact, mobile browsers can handle this page pretty well.

Here's what's wrong: the user, having come to this page on his or her phone, now has to fill out 12 fields, including a simple spam blocker, to RSVP for the event. As an end user who just loaded the page on the phone to RSVP as the ad suggests, I didn't expect I'd have to bloody my thumbs entering this much information (hyperbole anyone?). As a technical user, I was hoping for an easy way to add this event to my calendar and maybe directions to the college that aren't a PDF download.

Context (added July 1, 2011)

So @aardrian has a QR code on his shirt that doesn't work but I'm pretty sure it leads to Indonesian midget porn. #smdayBUF

In this example, I proudly wore my QR-code-embroidered shirt (well, one of them from my growing collection) to the Social Media Day event last night. Throughout the evening people asked about it and a few even scanned it — successfully.

As the night wore on and the bar became darker, there just wasn't enough light to get good enough contrast for a QR code scanner. Once that happened, people asked what it linked to and one, as evidenced by the tweet above, even made his own assumption.

Knowing how people will scan your QR code comes from understanding where they will be when scanning (a water park, the office, a car) and what they will be doing (drowning, working, driving). It's fair to assume that not all scenarios will be a good fit for a QR code, including some obvious ones like billboards on highways, so make an effort to understand the context of where it will appear along with the value of the reward for making the user go to the trouble.


These are examples of good intentions. However, lack of a solid business case leading up to putting the QR code into advertisements, compounded by lack of understanding the medium after the QR code is scanned can result in a negative experience. Pile on the clear lack of testing on mobile devices for the first one and you have a recipe for disaster.

Now, to connect this with social media (because that's where I am borrowing these rules and where this discussion started), if you don't understand the medium, don't have a business case, and aren't prepared to commit (whether by testing or engaging) then that tweet way above is spot on — forget about QR codes.


Tuesday, June 28, 2011

A Little More on Klout and My Magical Gift

Change in my Klout score over 30 days

The graphic above shows my Klout score as of today. Not only did my number jump dramatically, I also shifted from being an Explorer to being a Specialist. That corresponds to a two column jump in the Klout 4×4 graph of social influence. I don't know how that is measured, though, so I can't exactly tear that apart.

Change in my Klout score over 30 days

This graphic shows my Klout score as of yesterday. A balmy 49, or a 6 point jump from my 42 from one week ago. You'll notice both graphics indicate a jump of 2 points in the last thirty days, though without clarification of whether that's measured from my score on day 0, or my change from my last score. So I revisited my Score Analysis:

Chart of my score over 30 days.

As you can see in this chart, the lowest point of my last thirty days is still above 50 (the tooltip in the image represents the point at the lowest part of the graph, coming in at 50.39). If you look at my score from yesterday, I was below 50, and yet according to this chart not only have I not been below 50 in thirty days, yesterday my score was 53.08. And to push this point again, precisely one week ago, my score was 42, as I proudly displayed in this image:

Klout score bubble.

I have jumped 11 points in seven days. I have done it magically since it's not even tracked on the chart. I have apparently developed magical powers over time, space, and social influence.

This demonstrates precisely why I am suspect of any social influence scoring model that is used by any organization to deliver any level of different service or benefit.


Monday, June 27, 2011

Social Scoring As the New SEO

Change in my Klout score over 30 days

Lately I have noticed that Klout is getting a lot of traction in discussions about social media. It may be that there is just more coverage, or the name has started to penetrate to more users, or the idea of social scoring is becoming more interesting to marketers. It's also possible that I am keyed into it because I don't really believe the scoring model is truly accurate. On this last point, you can see my blog post from Tuesday, Does Your Klout Score Mean Anything? to understand my skepticism.

My skepticism was not assuaged when I noticed that my Klout dashboard (yes, I finally signed up) claims that my score has gone up 2 points in the last 30 days, justified by the handy plot Klout provides of that same period, even though I know 30 days ago my score was 7 points lower (I took a screen capture). It appears that after I added my Facebook and LinkedIn profiles to Klout, it retroactively adjusted my score. Which sounds very much like revisionist history and further erodes my confidence in Klout's platform.

Despite this clearly math-free scoring model, Mashable opened an article (with the technically accurate, but still unfortunate page address of klout-gate) last week with this unsurprising statement:

Your Facebook influence, as measured by Klout, will determine your access level to select brand pages on Facebook — and it could net you perks.

This weekend The New York Times published Got Twitter? You've Been Scored where it examines the scoring trend and what it means for consumers.

Companies with names like Klout, PeerIndex and Twitter Grader are in the process of scoring millions, eventually billions, of people on their level of influence [...] [T]hey are beginning to measure influence in more nuanced ways, and posting their judgments — in the form of a score — online.

The New York Times article takes it a bit further and outlines specific examples in the real world where a Klout score can have an impact on the perks, benefits and discounts a consumer might receive from brands:

More than 2,500 companies are using Klout's data. Last week, Klout revealed that Audi would begin offering promotions to Facebook users based on their Klout score. Last year, Virgin America used the company to offer highly rated influencers in Toronto free round-trip flights to San Francisco or Los Angeles. In Las Vegas, the Palms Hotel and Casino is using Klout data to give highly rated guests an upgrade or tickets to Cirque du Soleil.

While I believe this has significant parallels to SEO in its early days, the key difference is that this impacts individual consumers, not organizations trying to promote their web sites. The direct correlation with a tangible cost (whether by discount or freebie) also appeals to an individual more readily, since you don't need an accountant and a balance sheet to figure out the benefits.

Since Klout score is partly driven by your Twitter follower count, Twitter replies, Twitter retweets, Facebook friends and comments, and now LinkedIn activity (with the promise of Foursquare in the future), we can expect to see the recent scourge of Twitter follower guarantees (among others) make a comeback. Within a company, you might have an individual tasked with wading through these detrimental sales pitches, but the average consumer who wants a discount at a hotel might not understand that a promise of thousands of Twitter followers may ultimately guarantee a significant reduction in one's "score."

If you read this blog, then you are either technically capable, fancy yourself a social media expert, or think you know something about SEO (or you're my mom — Hi Mom!). Now would be a good time to take those skills and try to help your friends and family. Look for the same trends we saw in SEO spam ("Submit your site to hundreds of search engines!") and social media spam ("Twitter followers guaranteed!"), only with promises tweaked to the new target market (consumers).

Just as some companies wear their Google Page Rank with pride (or shame), we may start to see individual people do the same with the per-person analogue — their social influence score.

Wednesday, June 22, 2011

The Logo Using Only CSS


<p id="evolt">
<a href=""></a><br>
<a href=""></a><br>
<a href=""></a>

Begat this:

See a little more detail and the CSS over at

Tuesday, June 21, 2011

Does Your Klout Score Mean Anything?

Klout score bubble. I tend to be wary of anything that reviews what I write and offers to spit out a score of how valuable it is. Just as I have mistrust for web sites that claim to be "Bobby Validated" for accessibility (I'm dating myself a bit there), or personality tests that tell plot me on a chart based on answers to a small set of questions (which has been co-opted by eHarmony), or guaranteed Google/SEO ranking offers (insert any spam you have gotten through your web site's contact form).

I've commented before on how, in social media, the number of Facebook likes you have or Twitter followers (Lots of Twitter Followers Guarantees... Nothing) you've amassed really doesn't say a lot about your ability (having the word "guru" in your Twitter bio is an automatic disqualification).

Of course it should come as no surprise to those who know me that I don't consider a service like Klout ("The Standard for Influence") to really do a good job of quantifying how good or bad, influential or not I am at social media. I have no illusions about my influence #8212; I have less than 400 followers on Twitter, I don't use Facebook for anything other than crazy talk, I hardly check my LinkedIn account, and I'm not exactly breaking new ground on this blog. My Klout score should be low. But I'd also like to know how it's generated and if it translates to anything tangible.

Cue an interesting article earlier this month, Why your Klout score is meaningless. The author claims to have a PhD in statistics, which admittedly makes no sense to me (I get natural math, not the invented math of economics or statistics), but puts him well above me to evaluate whether or not Klout has any merit. His conclusions:

It often does not correctly order individuals in terms of how influential they are, is easy to game higher simply by adding a Facebook account, and does not respect some very basic monotonicity rules. Put simply, it acts like a derived measurement.

Earlier in the post he offers research has repeatedly shown derived measurements to be inconsistent and not trustworthy individually.

The Director of Ranking (interesting title) from Klout responded in the comments provides some context for how Klout works, but doesn't explicitly challenge the core assertions of the article.

A professor of statistics and political science and director of the Applied Statistics Center at Columbia University weighs in on the previous discussion with his own blog post, also titled Why your Klout score is meaningless. He doesn't dispute points from the original post, instead taking issue with the application of the term "derived measurement." He does have this to say about Klout:

The problem with the Klout score is not that it's subjective but that it's cloudy: we don't know what it is. To understand a cloudy measurement, one has to poke it from the outside. [...] Which makes sense given that Klout itself seems like a tool for ... selling itself! Sort of [like] other notorious rating schemes such as the Places Rated Almanac and the U.S. News college ratings.

For a bit more vitriol wrapped around a play-by-play between Klout and others, you can catch up in this post from November, Klout. Nail. Coffin. Who cares?.

Since these posts were written Klout has moved to add two more features that will likely further affect a person's Klout score, and further confuse (or confound) those trying to measure it. Klout introduced a "+K" button, which is a take-off of Google's own derivative "+1" button, allowing people to indicate that a particular person has influenced them on a topic (Introducing +K: Your Influence is Topical). In addition, Klout will now allow you to associate your LinkedIn account and factor that into your overall Klout score (Measure Your LinkedIn Klout)

If you aren't concerned about how Klout and other arbitrary measurements (such as your number of Twitter followers) can affect you and others, consider these examples (from Get ready. Social scoring will change your life.):

  • The Palms Hotel in Las Vegas is providing perks to guests based on their Klout score (an assessment of social media influence)**
  • By the end of the year, Twitter said their new analytics will provide influence scores for every user.
  • People are now curating lists of the most influential bloggers by Klout score.
  • Virgin Airlines offered free flights on a new route to people with high influence scores on Twitter.
  • Hoot Suite allows you to sort Twitter results by the influence of the people in the list.

There are already people and organizations who perform triage (at best) or just selectively respond to others based on their Klout score. Apply the same motivation for spammers to game the Google algorithm to Klout scores, and you may face the risk that your former ability to reach out to a brand using social media is hampered because you haven't paid for enough tweets in mainland China to artificially boost your ranking (or that those getting the responses did).

I think my next career might lie in goldfarming followers, likes and re-tweets.

Monday, June 20, 2011

Make Your Own TLD? (I want .bacon)


The Way We Were

Years ago the general public was aware of three primary generic top level domain extensions (gTLD): .com, .net and .org. There was a huge "land" rush as the dot-com bubble grew and organizations were willing to spend absurd amounts of money to get the .com extension for their business, even building the massive cash outlays into their business plans. Then search engine marketers (snake-oil salesmen) and web development gurus (hacks) started to push organizations to get the .org and .net versions of a domain for assorted reasons.

As ICANN moved to provide more gTLDs, many of which didn't make sense to the average user and gained little traction, we saw extensions like .aero and .museum appear. Registrars hopped on the bandwagon and promoted every new TLD as a requirement to branding your site.

Then countries got into the business by selling access to their ccTLDs, such as Tuvaloo with .tv marketed to to the television industry, and with enough advertising behind it, users started to accept them. The explosion of Twitter and the need for shortened web page addresses further promoted otherwise ignored TLDs, and services like helped put the Libyan TLD on the map. Some of us are still curious to see how .xxx pans out.

How It Has Changed Again

As of today, ICANN has approved a process to allow organizations to apply for a TLD of their own (read the PDF press release [and you thought UDRP was complex]). If you are Google, for example, you might want .google. Apple might want .apple. It is conceivable that you could see addresses for,,,, and other unlikely but possible combinations. Apparently the 22 gTLDs already in play don't offer enough variance for the web.

The process itself may be relatively straightforward. ICANN will make applications available (get the May 2011 draft of the Applicant Guidebook) from January 12, 2012 through April 12, 2012, giving companies time to develop a marketing plan and come up with justification to pay the $185,000 application fee and, if approved, the annual $25,000 fee.

With those numbers you can see the potential for a small set of players who want to give it a go. From ICANN:

New gTLDs will change the way people find information on the Internet and how businesses plan and structure their online presence. Internet address names will be able to end with almost any word in any language, offering organizations around the world the opportunity to market their brand, products, community or cause in new and innovative ways.

If my favorite "cause" ponies up that kind of money for a TLD, I can assure you I'll re-evaluate whether that cause really needs my money.

Given how inexpensive it is to obtain a domain name using one of the current TLDs, I am not sure how ICANN expects companies to justify the expense. The limited window implies that this is just an experiment, but is probably also designed to get organizations to move before they lose their chance. Whether or not the public will understand, and use, these is a different story. Instead, I see value in an existing company (with some cash) to consider finding a two-character gTLD, that is not already assigned through the ccTLDs, and rolling its URL own shortener service.

What is also not clear is what happens when two valid organizations in different spaces apply for the same gTLD. If Champion (spark plug makers) and Champion (t-shirt makers) apply for .champion, how is that sorted out and is the application fee lost for the loser of that decision? Here's ICANN's answer from the FAQ:

It is not feasible for two or more identical strings to occupy the Internet space. Each name must be unique. If there are two or more applications for the same string (or confusingly similar strings), the String Contention procedures would come into effect. Refer to module 4 of the Applicant Guidebook for more detailed information regarding the String Contention procedure.

Module 4 defines two methods in the String Confusion Dispute Resolution process to address this (with the presumption that the parties at odds with one another weren't able to sort it out on their own):

  1. Community priority evaluation,
  2. Auction

The first one only applies to community-based organizations and an unspecified deposit is required to participate. The second one, the auction, isn't allowed when the extension is for a geographic name. The document then goes into detail outlining the concept of an auction along with general rules (currency, defaulting, etc). At that point, it truly is a pay-to-play scenario.


Given how many users still type a web page address into Google search, will the new gTLDs really matter? I'm not sure I understand the problem hat ICANN thinks it is solving, or the business case to justify the purchase, but I am also not privy to the players on the board or the pressure they might be getting elsewhere.

The video below shows the vote — well, it shows the vote for the change, but pans away so you don't see how many voted against this or abstained (13 for, 1 opposed, 2 abstentions).


Update: 10:15pm

Mashable, a resource I generally consider good for quickly covering stories but not so good for providing much depth, has taken some time today to review the ICANN guidebook for the new gTLDs and put together the post 9 Things You Need to Know About ICANN’s New Top Level Domains. It's not terribly detailed, but it does provide a good, quick overview if you need to know something right now. You know, because January is right around the corner.

Update: June 13, 2012

ICANN has announced the list of requested gTLDs. I provide links and list some my favorites at the new post ICANN Announces Requested gTLDs.

Update: May 8, 2013

The marketing manager for guest-writes a post at .net Magazine ("Google sets precedent for new gTLDs to be open") detailing how Google may be opening up any of the new gTLDs it acquires, instead of restricting their use to just Google brands. The writer hopes others will follow Google's lead.

Tuesday, June 7, 2011

CSS 2.1 is Finally Final

W3C It's only been 13 years, but CSS version 2.1 is now officially a W3C Recommendation — essentially meaning the specification is final. Which of course means you are now all free to use it in your web pages.

CSS2 became a W3C Recommendation on May 12, 1998, over 13 years ago. Since then the CSS Working Group has been developing CSS Level 2 Revision 1 (CSS 2.1) to correct errors and omissions from the original CSS2 specification. For context, the HTML 4 specification was approved on April 24, 1998. The HTML 4.01 specification was approved on December 24, 1999. Between those two versions, little more than a year and a half passed.

The W3C has a slightly confusing progression of a specification before it is considered a "standard." There are essentially four steps:

  1. Working Draft (WD): This is the first time a proposed specification is shown to the public and open for comment.
  2. Candidate Recommendation (CR): Significant features are mostly locked and feedback is requested in how to implement the standard.
  3. Proposed Recommendation (PR): The specification has been submitted to the W3C Advisory Council for approval. Changes at this point are rare.
  4. W3C Recommendation (REC): The specification is final and endorsed by the W3C. This is what the general public considers a final standard.

In July of last year I wrote up a post about the CSS 2.1 specification status ( CSS 2.1 Still Not Final). Bruce Lawson picked up on the irony in my post (or at least my bewildered expression) and used humor to convey just how this hasn't stopped anyone from implementing it for the last decade-plus (CSS 2.1 “not ready for use” says journalist).

The key reason that this is newsworthy is because CSS level 3 (CSS3) is relying on CSS 2.1 to be wrapped up before it can move forward to the final steps. The CSS Working Group acknowledges that future CSS specifications rely on this final step (read the original release from June 30, 2010: An Update on CSS 2.1):

[CSS] 2.1 must be released as a Web Standard because that's one of the current cornerstones of the architecture of the World Wide Web. We cannot make the next steps, CSS 3 even module by module, happen without 2.1 before.

In case you want to replace the last version of the CSS 2.1 specification that you have hanging on your refrigerator, you can grab the latest (today's) version at the W3C site:

Just to make sure CSS3 isn't lost in the certain global buzz over CSS 2.1, the W3C pushed two more CSS specifications to W3C Recommendation status today:

For further reading, visit the press at the W3C site:

Update: June 10, 2011

Net Magazine has a brief piece on the wrap-up of CSS 2.1 by interviewing some known standardistas and getting their thoughts on the milestone approach versus the constant iteration approach: It's official: W3C finalises CSS 2.1

Monday, June 6, 2011

Testing IE Versions via IE Compatibility Modes

Internet Explorer logo This past week I have encountered people asking about testing for Internet Explorer browser versions in real life, on Twitter, via email, and spray-painted under a bridge (along with the phrase I don't want the world, I just want your half). I have seen response after response directing web developers to simply use the different browser modes in Internet Explorer 9 for browser testing, in particular addressing those who are using Macs and do not want to install multiple versions of Windows for each browser instance.

I whipped up some HTML and CSS of the most basic (and invalid) sort to run my installations of Internet Explorer through the paces. Historically I have found subtle differences between the emulation modes and how the original version of a browser (specifically IE) might render a page. This time out I simply found the emulation mode testing process so confusing that I suspect many developers who rely on this method are doing it incorrectly.

Four Versions of Internet Explorer

These images represent screen shots from Internet Explorer versions 6, 7, 8 and 9 — the full install version of each, using default settings, no emulators. For the most part we can see differences between IE6, IE7 and IE8, while the differences between IE8 and IE9 are negligible.

Screen shot in IE6.
Internet Explorer 6

Screen shot in IE7.
Internet Explorer 7

Screen shot in IE8.
Internet Explorer 8

Screen shot in IE9.
Internet Explorer 9

You can see that IE6 has trouble with the 24-bit transparent PNG files, and both IE6 and IE7 apply the padding to the wrapper element differently than IE8 and IE9 do. IE6 is also not setting the height of the element that is the green box, but from IE7 onward the height declaration is honored.

Internet Explorer 9 as an Emulator

The problem is that once you install IE9 you are kind of stuck. Unless you are willing to spin up a virtual machine for the last few versions of IE you want to test (and I still recommend you at least see how IE6 will munge your pages), you have to rely on emulators. Many developers think that IE9 will allow them to see how older versions of the browser saw pages, and to some extent it will. The confusing part is that not only is there a Browser Mode setting, there is also a Document Mode setting. Using combinations of those two settings, I generated the following screen shots:

creen shot in IE9. Browser Mode: IE7. Document Mode: IE7.
Internet Explorer 9. Browser Mode: IE7. Document Mode: IE7.

Screen shot in IE9. Browser Mode: IE8. Document Mode: IE7.
Internet Explorer 9. Browser Mode: IE8. Document Mode: IE7.

Screen shot in IE9. Browser Mode: IE8. Document Mode: IE8.
Internet Explorer 9. Browser Mode: IE8. Document Mode: IE8.

But those aren't all the modes. You can also run the browser in the following slightly confusing modes. You might notice that "Quirks Mode" most closely resembles the rendering in IE6 above, without the issues displaying PNG files. This is as close as you can get to testing IE6, and even that doesn't do justice to some of the rendering issues of that rather old browser.

Screen shot in IE9: Quirks Mode.
Internet Explorer 9: Quirks Mode.

Screen shot in IE9 Standards Document Mode
Internet Explorer 9: IE9 Standards Document Mode

Assuming you simply toggle the "Compatibility Mode" in IE9, which of all the above combinations do you think you'll get? Or better yet, how do you think your page will render? This experiment shows that it most closely resembles the rendering used for its IE7 browser and document modes. I would have expected it to use IE8, or at least some clear indication of which browser version it is trying to emulate. This example alone shows that even a quick cross-browser test in IE9 to test older versions of IE may provide unexpected results.

Screen shot in IE9: Compatibility Mode.
Internet Explorer 9: Compatibility Mode.

Even if you understand all the document and browser modes available in the IE developer tools, that doesn't mean your web staff, client, or random web savvy user will. The best way to truly test how a site renders in older versions of a browser is not by running emulators, but by running the versions of those browsers themselves.

If you still insist on going down the path of testing using emulators, you should take a few minutes to read the Microsoft article, Testing sites with Browser Mode vs. Doc Mode. The article outlines what each of those settings means and how it adjusts the rendering of the page.

The article makes an assumption, however, that as a developer you will want to serve up different doctypes and meta tags based on how the browser reports itself in the user agent string as defined by the Browser Mode. This image illustrates that:

Flow chart showing IE's versioning and compatibility system.

If you do have a site that adjusts itself based on the browser, given the extra steps to configure the browser, test and compare differences, it's still probably easier and definitely fool-proof to use the specific version of IE that you want to test. This way you don't have to worry about incorrectly read UA strings or mis-selected settings in your browser emulator.


Suck it up and install some virtual machines to get each version of Internet Explorer for testing.

Update: February 4, 2013

Microsoft has made virtual machines with old IE versions available to developers at its Modern.IE site. There really is no excuse now to claim you can't test in older versions of Internet Explorer.

Update: March 14, 2013

Typekit points out the dangers of cross-browser testing with IE9′s Browser Modes when it comes to font rendering. Typekit comes to the same conclusion as I do: Instead, you should test with real installed copies of IE7 and IE8[.]

Update: March 15, 2013

Found this post on Stack Overflow, dated just a couple months before I wrote this post: How well does IE7/8 mode in IE9 compare to actually running IE7/8? Many specific issues there.