Skilled modeling is key to efficient problem solving.
At Socialesque we have a rather unique approach to new media analytics that we refer to as our “Rules of Engagement.” Basically, we begin by working with a customer to identify the expected, desired, and undesirable behaviors of users of the content — a step we call Modeling.
Once we have a good model of the problem the customer is trying to solve and/or the behaviors that they are really interested in exploring, we turn to Instrumentation, or the task of identifying the numbers and metrics that we need to collect in order to enable this particular analysis. In some cases the customer already has an existing database, so Instrumentation becomes a processing of adapting their existing database schema to one that will be required for the later analysis and visualization.
In the Analysis stage, we begin to look at the numbers themselves in order to see if our initial assumptions were correct and/or if we’ll need to adjust the instrumentation in order to gain the desired insights.
Finally, we use the analysis in order to both improve our model of player behavior and to look for opportunities to tweak the application in order to enhance overall engagement — a step we call Refining. All four stages are important to extracting actionable insights from the system and putting them to work to make the content more engaging and improve a customer’s bottom-line.
Perhaps the secret weapon in all this, however, is the first step — modeling the problem space. In the AI community, it is well established that “having the right point of view,” “casting the problem in the right form,” and “conceptualizing a problem correctly” makes all the difference in efficient problem solving (See Saul Amarel’s article in IEEE Spectrum, v. 3, April 1966, pp. 112-114 for a good explanation).
For example, the classic “heuristic search” approach to problem solving involves an initial step of representing the problem mathematically or symbolically (a.k.a. ‘modeling’), followed by the cataloging of potential solution steps, followed by a search for the optimal (or sometimes just satisfactory) solution among all the possible options. In fact, Amarel (1966) argues that it is the selection of the model that demands the most creativity in this process of problem solving. This just makes sense; a bad formulation of the problem can lead to an endless (or fruitless) search.
So a key first step, it seems, would be the creation of an effective model. Scholars like Amarel suggest that powerful imagery, abstraction to appropriate spaces, flexible associations, and rich generation of analogies are the key elements of the creative process of Modeling. This type of skill isn’t gained overnight, however. It requires exposure to many different frameworks and models, as well as knowledge of previously successful or failed solutions; in short, it requires expertise with analyzing engagement.
So even if you have a robust system for logging, and a good visualization and querying toolkit, you might find yourself staring blankly at Pie Charts if you don’t begin with a good model. And the surest way to get there, is to talk to the experts at Socialesque.
May 28th, 2009 at 6:13 pm
hey this is a very interesting article!