Desktop Analytics Platform: Still Alive?
The wooden doors of the saloon have long since fallen off their hinges.
A sun-bleached steer skull gleams under the hot desert sky.
And tumbleweed creeps across the road, bumping against the old hitching post.
It’s been a ghost town around here lately, since I posted my last message in June, and the lost signs of life have not gone unnoticed a few of you out there.
Three or four times a month, I get an email from someone curious about my Desktop Analytics project, who maybe saw the GUI mockup, and wondered whether I was still around and still actively working on an implementation.
As it turns out, I’m still here. And I’m still working hard.
In fact, the ongoing emails from interested bystanders have really kept me motivated and have reaffirmed my confidence in the strength of the concept. As my way of saying thank you to the people who have sent me their questions, I’ll take a little time tonight to give you a long-overdue status report. Then, sometime within the next few days, I’ll post a sneak peak into the architecture I’m developing and some of the cool features that I hope to unleash on my users early next year.
I’m thinking…mid to late Q1…
First, a quick refresher, for those of you who might not have been around when I originally floated this idea.
I’m building an application analytics platform, analogous in many ways to Google Analytics or Omniture. But those solutions focus on analyzing web traffic to answer questions like “which search engine keywords provide me with the most traffic?” and “which of my content resources have the highest click-conversion ratio?” Those types of metrics were invented my the advertising industry for messuring consumer responses to ad campaigns. Long before Google emerged from the labs at Stanford, ad executives asked Nielsen analysts for their conversion numbers.
The web analytics software on the market today still reflects that mentality: the internet is a medium for delivering content. Some of that content attracts a lot of eyeballs. Some of those eyeballs tell their mouse-fingers to clicky on an ad. And some of ad-clickers buy hilarious cat calendars from Amazon.com.
Web analytics is all about maximizing throughput of the content/eyeball/mouse-finger/cat-calendar pipeline.
Developers of applications, rather than publishers of web content or cat calendars, get short shrift in the whole equation. Sure, we can get detailed statistics about which types of web surfers are more likely to download our products. But when the last byte finishes downloading, our visibility into that user’s behavior evaporates.
I want an analytics package that tells me how users actually interact with my software.
How many people are using that new spellcheck feature I spent so much time developing? How often do my users fire up the application? Every day? Once a week? Do they use the software all day long, or just for a few minutes in the morning? How many menu clicks does it take to find certain features? Are people getting lost in my user interface? Do they invoke the HELP system? Do the “Premium Edition” users actually use any of the premium features, or do their behavior patterns look just like the “Free Edition” users? How many people are still using a legacy version, and which features should I improve, to motivate an upgrade decision among those users?
Answering questions like that will help me develop better software.
And I like to think that “better software” is almost as important as “good keywords” in driving new revenue.
Anyhoo…I hope I’ve gotten you a little sweaty and breathless…
Now: status report. Tomorrow: architecture details.
The software comes in three pieces: an embeddable library, a server, and a reporting client.
The embeddable library is a tiny piece of code that developers will link into their own “host” applications. The code is written in a snazzy little programming language called D, which compiles to native machine code, providing C linkage on Windows and on x86 Linux and OSX. No matter what language or platform you’ve used to write your application, as long as it provides some mechanism for linking with C libraries, you’ll be able to embed the analytics library in your code.
As of today, the Windows version of the embeddable library is about 95% complete (the OSX and Linux versions are about 75% done).
The server is implemented in Java, so you can run it pretty much anywhere (I personally recommend a “Mr Coffee” deployment environment). The server is about 85% finished, and can handle about 100 requests per second on my dinky little 2-year old laptop. It doesn’t use Tomcat or JBoss, and it doesn’t depend on a gazillion application frameworks. It’s a standalone HTTP server, talking to a MySQL backend, and you can get it up and running within minutes (hint: no XML configuration).
Finally, there’s the reporting client, with all the sexy charts and graphs. I’m only about 10% done here, after making a bold decision to switch my development environment. I started out working in C# with WPF/XAML, but I recently switched to Flash/Flex/AIR (for reasons that I’ll describe in more detail later) and now I’m playing catch-up.
That’s all for now :o)
I’ll be back soon with more details. But right now, I just wanted to check in with you guys, let you know that the project is still rolling forward, and that I’m anticipating a Q109 release.
January 7th, 2009 at 5:07 am
I like this - except the C part, but that’s just because my own little app I could use this in is written in Java and I can’t justify buying a Mac just to link to a lib.
P. S.: I had to retype my whole comment because I forgot to type in the captcha and the comment text was erased after submitting.