Thursday, May 31, 2012

More on doing what one is inspected to do instead of what one is expected to do


I have already written last year about the seminal phrase of Lou Gerstner cited in the title of this post,  about the negative side of addiction to metrics as behavior controllers.      Today I found a blog posting by one of my IBM research colleagues, Matthias Kaiserwerth, the director of the Zurich Research Lab, entitled: Big data analytics - why delayed flight makes for more satisfied customers.    In this posting  Matthias quotes a talk by Andy Neely that indicates some of the phenomena associated with metrics.   

One example is that when a measure in a call center is to resolve a client's issue within 2 minutes, the agent starts to close the call after 1:45 minutes, regardless if the client' issue is resolved, since the agent is measured on length of the call and not on results.   Actually I saw such cases in which some administrators who are measured on time of handling requests preferred to find some reason to close the request and not handle it since the time exceeded their metrics.  

Another example that landed its name to the cited posting 


One of the major  metrics for airlines is "friendly staff",  using analytics techniques the results show that their is a positive correlation between flights being delayed and the friendliness of the staff, since the staff is very nice in trying to calm down angry passengers that become nervous about missing flights (and sometimes promise things that their colleagues in the destinations don't really accomplish),   taking this ad absurdum - the flight company should deliberately delay flight to enable the staff to maximize their friendliness metrics.

Neely has his own version of Gerstner's phrase:  doing the right things rather then doing things right.  

As I noted in my original post --  the metrics oriented culture is in many cases damaging and we should go back to the basics --- understand what are the right things.

More - later

Sunday, May 27, 2012

Event processing and mobile devices

The current big buzzwords in the IT world are: analytics,big data, cloud and mobile.   These are believed to be the most influential trends on the IT directions.  I have already written about analytics and big data in some previous post, today I'll write on the mobile universe from the point of view of event processing.  


Putting the mobile devise symbol on a drawing of an event driven architecture, we can view multiple roles:
  1. Mobile device can be a producer of events -- this may be both implicit events, using the mobile devise as a sensor and tracking the location of the mobile device over time  or explicit events, where a human use mobile devise to report events (such as:  traffic jams, completion of tasks, start of the work day)
  2. Mobile device can also serve as consumer of events -- subscription to either raw or derived events can be manifested within notification in form of Email, SMS, or mobile application.   It seems that most uses of mobile devises are the consumer and producer roles.    There are also some more sophisticated possible uses, discussed next.
  3. One can build event processing agent logic using mobile devise, but it is more likely that the use in this case will be to use predefined templates in order to define specific instances of EPA. For example, if the application is subscription to events in my bank account, I might get a template giving me an opportunity to subscribe when my accounts reached a balance of more than X or less than Y, I received a deposit of more than Z, or that the total withdrawals for a single days are more than W.   Using the mobile device I can chose any of them and give value to the X, Y, Z, W variables.
  4. A dashboard showing the real-time state considering the streaming events can be also displayed as a mobile application
  5. Last but not least -- An EPA can reside on the mobile device itself,  collection data and doing some filtering or aggregation.
In a follow-up post I'll try to switch direction and look at event processing from the mobile point of view.
More - later.