There is a interesting post written by Romakanta in his blog, Datalligence. He has listed and described eight levels of reports that is prepared by BI Folks.
Here is Romakanta's list.
1. STANDARD REPORTS
2. AD HOC REPORTS
3. QUERY DRILLDOWN (OR OLAP)
4. ALERTS
5. STATISTICAL ANALYSIS
6. FORECASTING
7. PREDICTIVE MODELING
8. OPTIMIZATION
The area where I disagree with him is on calling everything reports. I would rather divide these things into various steps of Decision Management. First three comes under reporting, and rest come under Analysis.
Mostly the standard reports and ad-hoc reports are called reports, and the OLAP is more to dashboard than report. It gives a flexibility to play with the data and do some analysis.
Also, I can hardly make out any difference between forecasting and prediction. Prediction Modeling is a technique and forecasting is a business solution. Forecasting can be done in various ways like Predictive Modeling, Time Series (curve fit or waterfall), Multiple Modeling Approach, Business Rules Approach etc.
Optimization is again business solution to a business problem of low efficiency. It can be achieved by various methods including segmentation, Optimization Algorithms (Operations Research Methods like Transportation Problem, Simplex Method etc) etc.
Friday, November 21, 2008
Subscribe to:
Post Comments (Atom)
2 comments:
i mentioned eight level of analytics, not reports:-)
this was taken directly from the sascom magazine. i agree, the SAS explanation for predictive modelling & forecasting can be a bit confusing. but as far as i know, and the way most people refer to them, forecasting has the time factor in it unlike predictive modeling.
something like using prediction for continuous target variable, and classification for categorical target variable.
that's exactly why the analytics community need to come up with standard terms:-))
Romakanta,
Very true. The need for standardization is increasing. It is very hard to understand the terminologies found around.
Even the reports and techniques might have to be standardized. Everybody using different technique to solve the same problem or show the performance makes it is hard to work too. I am saying this in respect of the use of various metrics to show Model Performance. Few companies use divergence, others use KS; some use Market Richness Curve and other Lorenz Curve; some use gains table and gains chart while others use value add sheet. It creates lots of problems for both the consulting companies and the clients.
You have picket a valuable point. We may need to move towards that. Thanks for putting it.
-- Bhupendra
Post a Comment