Hi Guys,
If you are working in BI reporting you must have seen the formalas that are there for standard deviation, variance, Mean but we hardly used them as BI report developers so why are they there and who uses them was a question in my mind.
Reporting and Analytics are different things. A BI report developer is not a analyst and an Analyst does not have to know in detail about BI technology people can manage even with simple Excel .
So i know people use the term interchangibly and so its confusing what is the difference between the two.
Reporting
The navigation is structured and people know what they are looking for. Even adhoc queries by user come into reporting.
Generally we are reporting what is there. Like how much quantity was sold. We are not designing a model that will give a trend to predict what are the expected sales in years to come.
Analytics
Here analyst makes a hypothesis and uses data to confirm that hypothesis. It involves adhoc querying (dont confuse with adhoc reporting in BI). Generally data analyst role are separate from BI developers . Suppose business has a pressing question that can be answered by gut feeling and experience of managers still they can decide to go for data analysis and an Analyst can come up with a model to prove that the hypothesis made by manager is correct.
Now this requires analyst to prepare the model and fine tune it so that the analysis matches the hypothesis. In many cases it wont match which says the hypothesis was wrong.
Generally reporting is done on a structured data. Analytics can also be done but the chances for finding pattern in a unstructered data like web logs is more.
Famous tools for Analytics are - SPSS , SAS and R analytics.
I know Cognos offers SPSS from Cognos 10 onwards but to use it to business advantage you need a Data analyst or a developer with good knowledge of statistical analysis models and with ability to make hypothesis and test against data.
There are a lot of videos on statistical analysis on you tube. I would recommend you going through those videos irrespective of the tool . Because tool is just for helping analyst do their job. The thought process has to be within the person.
Analytics is best done on unstructured data. Like network traffic data for telecom company. Logs from the switches which can be loaded into a HDFS system ( Hadoop ) and analysed for patterns.
Difference between SAS and Cognos
SAS has a ETL part with it . ( i did not know that :-) )
some comments i picked from sites
SAS in my view is still fundamentally a data mining software vendor comparable to SPSS. Data mining sorts through data to identify patterns and establish relationships (Association, Sequence & path analysis, Classifications, Clustering, Forecasting).
The other vendors you mention are OLAP vendors (Cognos and BO are cube -based MOLAP solutions, MicroStrategy ROLAP). Online analytical processing (OLAP) for the most part allows users to derive information and business intelligence from data warehouse systems by providing tools for querying and analyzing the information in the Warehouse from different points-of-view. OLAP can be used for data mining or the discovery of previously undiscerned relationships between data items.
Generally, Cognos can be used as a data presentation layer that includes descriptive statistics and OLAP. SAS has a substantially more powerful ETL capability, so I would use SAS for data prep processes. Finally, SAS handles the inferential statistics and data mining, including predictive and optimization modeling.
To summarize, I've used SAS to integrate and perform ETL on data, Cognos to present he data, then SAS to analytically crunch the data. The analytical results may then be reported back through Cognos. This whole process can be made to feel
Big data and Analytics
With the coming of Hadoop we have the ability to analyse big data. Data like weblogs. Bio Informatics (DNA sequencing) . Online gaming data. Which user is likely to play which games. Geo mappings. These data can be multiple terabytes. So now analyst have a huge data set on which they can query (HSQL , Hive) and come up with predictive models or verify a hypothesis.
In dataware house the ability of analyst to run a hypothesis is less because data is structured and organised. So the changes that a Analyst can find a pattern in a unstructured data are more
If you are working in BI reporting you must have seen the formalas that are there for standard deviation, variance, Mean but we hardly used them as BI report developers so why are they there and who uses them was a question in my mind.
Reporting and Analytics are different things. A BI report developer is not a analyst and an Analyst does not have to know in detail about BI technology people can manage even with simple Excel .
So i know people use the term interchangibly and so its confusing what is the difference between the two.
Reporting
The navigation is structured and people know what they are looking for. Even adhoc queries by user come into reporting.
Generally we are reporting what is there. Like how much quantity was sold. We are not designing a model that will give a trend to predict what are the expected sales in years to come.
Analytics
Here analyst makes a hypothesis and uses data to confirm that hypothesis. It involves adhoc querying (dont confuse with adhoc reporting in BI). Generally data analyst role are separate from BI developers . Suppose business has a pressing question that can be answered by gut feeling and experience of managers still they can decide to go for data analysis and an Analyst can come up with a model to prove that the hypothesis made by manager is correct.
Now this requires analyst to prepare the model and fine tune it so that the analysis matches the hypothesis. In many cases it wont match which says the hypothesis was wrong.
Generally reporting is done on a structured data. Analytics can also be done but the chances for finding pattern in a unstructered data like web logs is more.
Famous tools for Analytics are - SPSS , SAS and R analytics.
I know Cognos offers SPSS from Cognos 10 onwards but to use it to business advantage you need a Data analyst or a developer with good knowledge of statistical analysis models and with ability to make hypothesis and test against data.
There are a lot of videos on statistical analysis on you tube. I would recommend you going through those videos irrespective of the tool . Because tool is just for helping analyst do their job. The thought process has to be within the person.
Analytics is best done on unstructured data. Like network traffic data for telecom company. Logs from the switches which can be loaded into a HDFS system ( Hadoop ) and analysed for patterns.
Difference between SAS and Cognos
SAS has a ETL part with it . ( i did not know that :-) )
some comments i picked from sites
SAS in my view is still fundamentally a data mining software vendor comparable to SPSS. Data mining sorts through data to identify patterns and establish relationships (Association, Sequence & path analysis, Classifications, Clustering, Forecasting).
The other vendors you mention are OLAP vendors (Cognos and BO are cube -based MOLAP solutions, MicroStrategy ROLAP). Online analytical processing (OLAP) for the most part allows users to derive information and business intelligence from data warehouse systems by providing tools for querying and analyzing the information in the Warehouse from different points-of-view. OLAP can be used for data mining or the discovery of previously undiscerned relationships between data items.
Generally, Cognos can be used as a data presentation layer that includes descriptive statistics and OLAP. SAS has a substantially more powerful ETL capability, so I would use SAS for data prep processes. Finally, SAS handles the inferential statistics and data mining, including predictive and optimization modeling.
To summarize, I've used SAS to integrate and perform ETL on data, Cognos to present he data, then SAS to analytically crunch the data. The analytical results may then be reported back through Cognos. This whole process can be made to feel
Big data and Analytics
With the coming of Hadoop we have the ability to analyse big data. Data like weblogs. Bio Informatics (DNA sequencing) . Online gaming data. Which user is likely to play which games. Geo mappings. These data can be multiple terabytes. So now analyst have a huge data set on which they can query (HSQL , Hive) and come up with predictive models or verify a hypothesis.
In dataware house the ability of analyst to run a hypothesis is less because data is structured and organised. So the changes that a Analyst can find a pattern in a unstructured data are more
Im no expert, but I believe you just made an excellent You certainly understand what youre speaking about, and I can truly get behind that.
ReplyDeleteSAS Training in Chennai|SAS Course in Chennai|SAS Institutes in Chennai