The Weather Channel's head of consumer application software argues that we have to take measurement on the desktop from Kafkaesque to truly useful.

I have felt for some time that our industry needs to develop a standard vocabulary for the measurement of desktop applications. Applications create a very distinct relationship with users, far different than websites. The usage is different; the value of that usage is different, and our metrics need to reflect these differences. This conviction was cemented during a Kafkaesque conversation I recently had with one of the measurement companies about a desktop application and the definition of a user. 

Are you sitting down?

One of our apps, Desktop Weather, has several options when launched. At one end of the spectrum it launches full screen and in focus (the application is in front of all other windows). At the other end, it launches, but then it only displays the temperature in the system tray. In either case, to me, I have an active user for that day. This is an application, after all, and both options display meaningful data to the user. 

The measurement company disagreed. They do not count a person as a "Unique Visitor" if the application only displays data in the system tray. Why?

"The user did not launch the browser."

What browser, I asked?

"The application," they corrected.

No, says I, the application launched and it is displaying data in the system tray.

"But it could have automatically launched at start-up."

And by doing so, I protest, I have a unique active user for that day.

"Visitor."

What?

"Unique Visitor."

But they are not visiting -- the application is on their hard drive. 

"That's what we call it and besides, they didn't view a page."

What page? This is an application screen. 

And on, and on. 

My point is not that the measurement company was wrong. The problem is that the industry uncomfortably uses web metrics on desktop usage. Right now, when we discuss user behavior, every concept we use is tied to the browser-and-page construct and superimposed on the application. This confusion leads to the combination of both application use and content usage, regardless of what the application is doing.

This lack of vocabulary is a problem with desktop applications and will be an even bigger issue when IP convergence allows all media to be measured (as a smart colleague mused the other day). With time/place shifting, multiple devices sharing home networked media and TiVo'd Yahoo!, we need to begin a dialogue around expanding the standard lexicon of measurement.

We need to start thinking outside of web.

I would argue that desktop apps make up a perfect first case study.

A good start? Define a user more accurately

Let's start by expanding the idea of an active user. For the sake of simplicity, we will say the measurement of a desktop user begins with installation and ends at an uninstall. This equation is simple and unambiguous. Installations can certainly be measured, and uninstalls, while not completely accurate, tend to be close.   

Note that things get a bit more complex when we try to separate the action from the user. We would want to also measure "unique" installs/uninstalls, because users regularly install, uninstall and over-install, skewing numbers and percentages. By noting "unique" and "non-unique" in this and many other metrics, we account for multiple actions by the same user. Why not just count unique metrics? The delta between your non-unique and unique installs, for instance, can tell you a lot about whether your application is having problems. (I'll explore this in a future column.) 

Install/uninstall can be used to analyze behavior, but we don't use it to measure our active users. As I mention above, I define an active user as one who launches the application. My "Active User" launched the application (or allowed it to be launched), separate from viewing any of its content. What if we define the population of people who launch the application as "Unique Active Users"? 

By using this concept of a "Unique Active User," we clear up the ambiguity of the Unique Visitor when measuring desktop apps (although I still have issues with this term). A Unique Visitor to an application can be better defined as an individual who a) has the application installed, b) launches the application and c) bring the application to full, in-focus state. This should make my measurement colleagues happy.

Now that we have created this separate term and defined the Active User, it frees us to also define an Inactive Users (just try to conceive of an Inactive Unique Visitor). Inactivity is a good metric to know. It can be used as both a measure of an application's dormant potential and a good barometer of its perceived value by the user. 

Calculating inactive users requires defining a measurement period, such as a month, that has a starting Active User count and an ending Active User count. In this equation, uninstalls are always an absolute number:

Ending Active Users minus Starting Active Users minus New Users within the Month

 Plus

Uninstalled Users within the Month

 Equals

Inactive Users

As I mentioned before, there is a margin of error here as users may uninstall the application offline and therefore not report the event. This too can be accounted for by aging out inactive users after a reasonable period.

These are simple examples of how we might expand the measurement of the desktop application. I could go on, however, my purpose is not to write the lexicon, or impose my concepts, but start a dialogue around the need for new metrics. The point is that we have to differentiate between the values of an application and a web page. These values (and the value) extend beyond a simple "page view" or full-state screen view. By improving desktop measurement we can begin to take this difference into account.

Furthermore, publishers and providers can begin using these metrics as predictors of churn and lifetime value in a manner similar to telecommunications companies.

This type of analysis would help everyone better understand the desktop medium and its value to consumers and advertisers alike.  

Not a bad goal in a world where I can Sling a laptop to TiVo. 

Matthew de Ganon is vice president and general manager of consumer application software at The Weather Channel Interactive (TWCI).  For TWCi, Matthew is charged with leading the company's rapidly expanding efforts in the consumer applications space, with a focus on desktop software. Matthew oversees development and execution of product and new business strategies; works with TWCi's robust ad sales and database marketing teams to develop innovative advertising products; and creates extensive relationships and other creative distribution channels for user acquisition.

Since 1992, Matthew has advised corporations on the strategic use of online technology, including Morgan Stanley, Bristol-Myers Squibb, Hewlett Packard, NCR, LEXIS-NEXIS, Prudential Securities, ABB, Sony and WorldCom. His experience with such clients led him to architect W3 Organizational Modeling, a model for performance optimization and profitability improvement of online operations. From 1994 to its successful sale in 2001, Matthew's abilities helped propel the evolution of K2 Digital, one of the early Internet business strategy companies.  Under Matthew's leadership as president and later CEO, K2 was named to AdWeek's Top 25 Interactive Agencies and Crain's "New York's Top 15 Players" and attracted many Fortune 1000 clients.

Matthew was then recruited to become president for desktop application developer Arcavista Corporation, best known for its work with RCA and American Idol.  At Arcavista, Matthew helped to transform the company, its products and its branding, responsible for the development of the company's vision, flagship Communicator Platform product, and its general operations.

Matthew was recognized as one of "New York's 100 Top Internet industry Executives" by Silicon Alley Reporter, and dubbed an "Internet Economy Architect" by The Industry Standard. He began his career at NBC, Universal Studio's Motion Pictures Group and also worked as an agent in Los Angeles. He later became vice president of new media of SCS, a software developer serving clients such as Marvel Entertainment.