10.We have the best modelers in the world
9. And they don't clean data; they only build the best models
8. Not sure if your Marketing guys can appreciate what we do
7. We typically don't work with your IT folks
6. Do you maintain robust no-mail Controls?
5. We need atleast 50,000 observations
4. Our Model documentation is the best; typically 186 pages long
3. How are your demographic variables bucketed?
2. Our last response model had a 82% lift over random in the top 3 deciles
1. Unfortunately two of the most significant model variables were not available at implementation
The challenge I see is that most analytics-focused consultants get lost in technicals; failing to understand that in almost 100% of the cases, understanding the context, communicating the real-life impact of analytics to the business and ensuring that outcomes can be robustly implemented are much more important that getting that incremental 2% lift in model performance or using a slew of alternative statistical techniques. And that businesses rarely have the benefit of perfect information and foresight that most analytics-focused folks typically assume.
NEED HELP ?
November 27, 2009
Subscribe to:
Post Comments (Atom)
2 comments:
I think business sense and outcome of real-time impact is always missed and mislead/mis-read sometimes by 'Statistical models' often considered in the analytics world as 'key skill' to work/survive - most of the times the real picture gets lost in the complex model(rather complexed) which fails reveal the bigger picture primarily due to lack of integration with business needs and blindly going by the technical/statistical expertise
Through my limited years of analytical exposure I've realized that one can solve around 90% of business problems using advanced data analytics (RFM, Value Migration, Trend Analysis) and intermediate statistical modeling (Regression - OLS & Logistic, Decision Tree, Cluster Analysis, Factor Analysis, Simple Time-Series Forecasting). The beauty of these methodologies is that the logic is easy to explain to the partners too and that to me is very important. We should strive to use approaches that don't appear as black boxes to our end-customers. Only when they understand the concept would they be more willing to implement the results coming out of these concepts into their decision making. This is when it becomes more of a "pull" rather than a "push".
Post a Comment