SmartClient Stats - what does your 'typical' smartclient PC look like?

It’s been a few weeks since I released thoughtex, and thanks to some kind words from a few folk like Frank, Scott, Glav and Ryan a number of people have tried it out. One thing I built into thoughtex was some feature usage monitoring. After being inspired by the massive amounts of user session data that the office team collect, and as a big fan of the general principle of measuing and finding out rather than speculating I thought this was an absolutely necessary feature to build in. Although the number of times the users changed the sytle of a node in my pet project is probably not of great interest to you, dear readers there was one set of stats taht I did collect that might be of broader interest - client capabilities. Here are some graphs of various metrics.

Breakdown of Operating System Types
(this is naturally somewhat selective since WPF, the platform my app runs on is only available for Windows Vista, XP SP2 and Server 2K3) 

Number of Cores

Render Tier
(measured by WPF - a broad indication of the client graphics capabilities. 0 = worst, 2 = best)

Tablet PC?

Screen Size
(this is the available screen size of the primary monitor, so it excludes space taken up by the start bar, and doesn’t include the whole desktop size in multi-monitor setups)

Smallest
932x640
Average
1423x998
Largest
2561x1601

The key things I took away from this (other than I need to get a 4-core system with a 2561x1601 monitor) is that I was interested in how many people had upgraded to Vista. Given the audience who read Ryan/Frank’s blog this is probably not that surprising. I was also slightly surprised at just how dominant multi-core systems are now. A lot of the single-core systems were also reporting render tier 0, suggesting they don’t have a real graphics card. I wonder if these are virtual machines (since I believe VMs have both these characteristics). Also it seems I shouldn’t go nuts implementing tablet-centric features.

For those interested, the number of data points was between 1K and 10K. I know these numbers are probably heavily skewed by the population of people who tried out my app, but I’ll certainly keep an eye on these stats when designing features as more “non-alpha-geek“ users (hopefully) start using my application.

Comments

Tim W
So it phones home and you didn’t tell us? Nice, I’ll have to remember to get a better firewall that stops outbound traffic as well.
16/04/2007 9:38:00 AM
JosephT Cooney
Tim W - I was planning to make turning off the feautre usage monitoring an "opt out" thing once the application is out of beta. Do you think it is important do do this sooner?
16/04/2007 4:03:00 PM
Michael Griffiths
Alright - Thoughtex just replaced FreeMind for me. Excellent program; well done.

In terms of your usage statistics - they’re interesting, but given it’s only been a few weeks since your application has been released your pie charts are unsuprising. (Early adopters, etc). What would be more interesting is monitoring the change in operating system types over time: would there be an increase in XP users? etc.

I’d also speculate that a number of users have laptops, explaining the screen resolution statistics (and potentially Teir 0 information). I’m using a Tablet at the moment, for instance…

Just a suggestion for your program - I’d greatly appreciate greater keyboard accessibility. For me, this would be the #1 improvement you could make (incidentally, using Yahoo! Web Services to grab related information was an inspired, and very cool, move). You can copy Freemind’s implementation (which is rather good). Unless I’m wrong, there’s no way to create a new subnode from the keyboard, or a node at the same level.

Two long term goals would be (1) to add Ink capabilities (notation - drawing all over the mind map - would be the best implementation. How cool would it be to make a mindmap, then scribble all over it?) and (1) allow for greater export/save issues. PDF would be ideal, for example.

Nice work; great program.
16/04/2007 4:13:00 PM
Tim W
I don’t have a problem with it per se, just with the secrecy of it. I didn’t see any kind of privacy statement in the license, that’s all. One the one hand, finding out your demographic for you is great, on the other hand users releasing "private" information without their knowledge is not. Having an opt-out would be great for the paranoid out there, but adding a line about the reporting on the homepage/license would be better. Don’t get me wrong…great product, and good idea for demographics, but unfortunately we live in a litigious society. Just my 2 cents.
18/04/2007 11:11:00 AM
Joseph Cooney
Tim W - thanks for the suggestion. I don’t collect any personally identifiable information (not user name, machine name, IP address, country of origin - I got nothin’) but maybe a few lines added to the license would be a good idea. I was also thinking of opening up the information (so anyone could query it at any time) to make it more transparent. Useful? Interesting?
19/04/2007 3:51:00 PM