The Cloud

At heart, Google’s website optimizer just tells you which version of a landing page converts better

Sometimes, a tool is simple and elegant in purpose, but that purpose is obscured by the technical brush you have to clear to get to it. Google’s website optimizer is a case in point.

Google Website Optimizer is a tool for directing people to different versions of a web page and counting how many undertake the desired action on each.

That’s it for the core value proposition. End of story. The sentence I just wrote should be emblazened in large bold letters on the front page of the web optimizer tool.

Now, some might complain that this core value proposition leaves out some key elements of what the tool does and what distinguishes it in the market place. I’ll list those in the order of descending guilt I felt in leaving them out:

• Website optimizer uses statistical methods for determining the extent to which your results are not just due to random chance. Specifically, it tells you the proportion of times you can expect the winner in your test to really win. The fact that it also gives you raw count data means you are free to apply your own methods also.
• Google website optimizer is just a special case of Google Analytics, with the implication that a lot of your understanding of that tool can just transfer over.

Then, there are things, heavily emphasized in the website optimizer introductory material, that I don’t regret leaving out at all. I’m listing these in decreasing order of irritation I experienced wading through them:

• Google website optimizer has a wizard that guides you setting up your experiments. Truth is I had a love/hate relationship with this feature. It does give you a good understanding of the basic things you need to designate. However, the approach is extremely cookie cutter and inflexible. In particular, it requires you to pass tests in setting up any experiment that may run counter to what your real end goal is, necessitating you back out some of the things the tools requires before you actually run the tests. Google even recognizes that some of the wizard’s requirements can be contradictory for the tests you actually want to run when they tell you to ignore some of the wizard’s warnings in their section on Advanced A/B testng.
• The javascript code in web site optimizer is often redundant and leads to diminished maintainability, particularly on advanced web pages using ajax. Some of the website optimizer help files actually point you to code, based on a deprecated API, that will not work with the current version of website optimizer (here’s the main culprit from the FAQ, the help version of this FAQ inside the wizard contains the deprecated code).
• The tool overly focuses on experimental design by forcing you into it right at the outset. The problem is that you often cannot achieve the gold standard of full factorial experimental design on which the tool was originally based. There are two related issues: (1) Often enough your concepts are developed to the point where you have different candidate solutions but not the rigorous theoretical basis for a full factorial design; (2) Full factorial designs can require an impractical amount of data collection to yield statistically significant results (actually implying that you are not stating your theory at the right level of generality, suggesting a special case of reason 1 just listed).
• Initially, it is not clear that website optimizer just counts one conversion event per visitor. Website optimizer is built on the premise that web pages are designed so that people only have one real option for converting. That premise works fine in a number of ecommerce scenarios where you want the person to make an order or not. It’s more questionable where the person may have a number of graded options, many of which fall short of fully placing an order. In particular, website optimizer may be less suitable for affiliate marketers.

Books are more like web video than they are like MP3’s; rendering counts

I’ve been thirsting after Amazon’s Kindle but wondering about the value proposition. I spend a lot on technical books from places like O’Reilly and APress. It would be great not to have to lug those around or have my computer open to read them online (something I actually don’t enjoy).

For now, based on this post from O’Reilly, I’m going to pass. The Kindle simply does not render elements of technical books well such as tables and monospaced fonts. The situation seems so egregious that it’s apparent to me something must change. Why bother purchasing when you know change must be in the air? We’re still in the early days of ebooks.

The complicating factor for ebooks is that, unlike music which early on had the mp3 format, rendering matters. What I mean by this is that the processor actually has to have the power to execute a visual layout. With mp3 audio, the translation from source to sound is much more straight forward and requires much less processing power.

In this regard, books are more like web video, which faced a similar hurdle earlier this decade. What eventually made web video work was a universal format (flash) that does a decent job. Ebooks don’t have that yet, and if they did, I’m not sure the Kindle could render it.

There’s technology innovation, and there’s business model innovation; the two are different

Andrew Sullivan asks if Plastic Logic is producing a Kindle killer. This is the same kind of non-sensical question that has frequently been asked of the iPod and the Zune, two competing music players (note to readers, the iPod continues to dominate the music player market). People who ask this kind of question confuse technological and business model innovation.

In all media, the distribution model has essentially dictated the business model; i.e., the cost requirements and revenue opportunities. Companies that profitably innovate around the business model, in ways that their competitors cannot emulate, create business franchises for themselves that allow them to continue offering their goods and services.

Plastic Logic has developed a new lighter weight, more durable way of rendering e-ink for consumers. In no way, does this technological innovation compete with the business model innovation Amazon introduced with the Kindle.

Amazon’s Kindle offering essentially creates a whole new means of distributing publications like books and periodicals that just happens to feed into a device. The device itself is easily replaceable as illustrated by the mere 1 year gap between the original Kindle’s introduction and the arrival of the Kindle 2.

Maybe Plastic Logic will power the Kindle 3. It won’t replace the business franchise Amazon has developed.

Joining Communities & Connections

I was honored that Nancy White asked me to join Communities & Networks Connection. It’s a hub for collecting various writers on social networking and building communities on the web. I think that’s a tough topic. We see some signs of success. But what makes success? Here are a few other questions that I think are relevant:

• Does the community have to be built around another community that exists the real world?
• Is physical contact or at least physical proximity necessary?
• How does a network differ from a community?
• How do networks and communities reinforce each other? Or, do they?

Distribution channels created a whole slew of professions in media and publication that just won’t be there anymore

If you consider traditional media and publication in the late 20th century, it was all about creating products that had two features:

• They could be distributed economically in physical form. In all cases, this implied some sort of bundling together or length requirement. Otherwise, the cost to the consumer per item of value would just be too high. For instance, music came in CD collections of songs, not as singles, op ed pieces came bundled in a newspaper, and novels had to be well over 100 pages.
• Largely due to the need to physically distribute items, they could not be altered after the fact, except with great cost. For instance, papers would have to print retractions on scarce newspaper space; for music to be altered would require exchange of a physical good, essentially replicating the distribution cost of the original with no additional revenue.

As a result of these two factors, a whole slew of professions grew up. Music producers who could successfully put together whole albums of songs became in high demand; editors became extremely important as quality control agents in all forms of media.

Throw away the requirement of physical distribution, and those kinds of professions are either not required or take on a whole new role. In particular, in publishing, we’ve seen the rise of the blog. Initially, blogs were purely amateur operations, but in the past couple of years, they’ve become professional. I’ll point to politico and talking points memo as two examples. Both employ professional journalists formerly of main stream media. Read articles, and you’ll see that corrections are posted throughout the day. There is editorial but often inserted as side notes after publication. Finally, both are highly focused on politics and political machinations, not bundled with extraneous news on science and culture except as it relates to the main theme.

Amazon is the digital wal-mart to books and the kindle is just another step down that path

You might think that, at $359, the Kindle is a high end product, and it is. It’s really still for early adopters at this stage. But, it’s pricing strategy for content should strike fear in the hearts of content producers, starting right at the high end.

The major innovation of the Internet relative to digital products has been distribution. The change in economics has been so profound that products like print publications that once could only conceptually be distributed in digital form may soon be almost exclusively so.

Amazon arrived on the scene in the mid-1990s. Like many other “etailers” at the time, it took orders over the web, saving on order entry. The innovation that allowed Amazon to dominate the online book market was economic distribution of individual books within parameters that consumers found appealing. The time to arrival was timely enough, and the cost low enough. Amazon was able to offer lower prices by making volume purchases. Its investments in distribution facilities was high enough and based on sufficiently unique insight so as to act as a barrier to entry to other participants.

Prior to the Kindle, Amazon’s distribution might was sufficient to require well above the standard 50% discount that retailers demanded to get a book, particularly a new book, on the shelves.

Now, consider the Kindle. It represents a revolution in book and other print content delivery. Within 60 seconds of selection, the book is in your hands, without an Internet or computer connection. No muss, no fuss. You click it. It’s there.

Further, with the Kindle, prices are dramatically less. A $30 hardcover book from a major publisher goes for $10. That’s a 66% discount. Further, that discount is being applied to the publisher’s high end product, the hard cover book.

Sure, in the long run, publishers might be able to adjust their production model, but in the short term, they’re still producing many bound books and other print media.  They’re making their investment decisions on potential revenues associated with the old distribution model. It hurts to have a portion of the expected revenue that still must partially support this model cut away.

Depending on the Kindle’s penetration, many publishers at the high end and perhaps increasingly at the low end will have to adapt to Amazon’s distribution model or die. That’s exactly how it worked with Wal-Mart’s big box retail model.

I just realized I’m writing a blog about creating community with no there there or at least very little

This blog is about creating community from streams and distributed resources. As a result, it’s bare bones. I’m going to flow pieces in and flow them out. Here’s the upshot:

• We’re hosted on tumblr, a remarkable platform that allows me to have a custom domain for free. Tumblr is also very standards compliant, almost to a fault. Finally, tumblr has an incredible RESTful API.
• We’re extremely bare bones but will add services by tapping into non-tumblr resources, to wit:
  • Commenting is most likely to come from TypePad Connect, but as you can see, it’s not there yet. Right now, the only way to return fire is to tweet at me (@BudGibson) or send email to Bud @ MichiganInnovators.org.
  • Backlinks may come from Google if I can find the right javascript snippet.
  • Not sure if there will be tags.
  • We will have analytics and likely campaigns.
• The Cloud will not attempt to have its own community. This may be my excuse to take a foray into facebook, and we may link up there. Right now, the community is really twitter, and we’re optimized for that.

I’m having commenting because it seems there needs to be an easy, built-in feedback mechanism. Some people will just wander in. What I really want though is for commenting to all feed into community creation. That means all commenting will be off-loaded to external services that have their own communities. The idea is to get it all to flow here.

Twitter stats I would like

• I’d be interested in knowing who’s clicking on the URLs I embed in tweets, or at the very least which of my tweets they’re clicking on the link from.
• I’d like to know who is referring me followers. What was the path the new follower took to get to me?
• I’d like a list of people I’m following who are not following me. Does the person produce intereesting stuff, and are they making a difference? If so, I’ll likely be happy to follow without a follow back. I’d just like a concise list of people who fall in that category.

The rise of the micro services: Is the web becoming unix all over again?

The original idea in unix was to write small utilities that each did their job well and could then be chained together to do more complex tasks. The idea is deceptively simple. You think to yourself: I can just focus on writing small things. The rub comes in two places:

• Actually chaining the services together. The services have to produce compatible streams; otherwise it’s a no-go.
• Getting other people engaged. People’s ideas of how the world works are different, in many cases because their needs are different.

Unix had slow uptake to begin with, particularly in the consumer market place. But now, it has arguably taken over. Web servicing primarily occurs on unix, the Mac OS is unix, the iphone is unix, and Android, a mobile phone platform by Google, is unix.

By contrast, the web’s uptake has been meteoric. Further, rather than this micro-services idea, you might say it started out with the all-in-one idea. I’ll just serve you a whole web page. That page might be composed of many pieces, but you’ll only see it as one.

Ajax, a way of loading in part of a page at a time as you interact with a web site, began to put paid to that idea, but the original implementations of ajax largely still pulled data from the same source. You could sometimes mash parts of different ajax pages together, but it was often awkward, requiring workarounds.

I think we’ve now arrived at Ajax part 2. Services are starting to be designed with the idea that they will be combined with other services. Designers of these services are pairing them down to the bare essentials.  For instance:

• tumblr really just provides blog posting services and an api to make it easier to post and extract data. There’s no commenting.
• typepad connect really just provides blog commenting services. There’s no posting.

It’s fairly easy to combine these two. So, while blogging used to come as an all in one package, it’s been disaggregated. It’s sort of like unix all over again, just from the opposite direction.

Tumblr is Blogging 2.0

So, what I’m really impressed about with Tumblr is how much it gets right:

• A web based editor that produces semantic xhtml
• Service integration: Everyone is outsourcing their infrastructure, be it photos, microblogging, mashups, you name it. You can suck it all into Tumblr.
• A great and straightforward RESTful API.

BTW, it seems like I’m going to have to integrate a commenting service of some sort. It doesn’t come out of the box.

Talk about minimal. I kind of like it.