Without further ado:
- Analytics work involves funky predictions: No. Most of the work done by analytics service providers is data cleaning and data visualisation
- Complex algorithms yield better results: No. Regression (including its derivatives like logistic regression) is enough to cover 99% of analytics work
- Fast multi-core processors are needed for analytics: No. The simpler the tools one uses more stable the model will be
- One has to learn tools like Matlab to do analytics: No. Matlab for all its goodness is not appropriate for analytics. There are many more simple free plugins/software available for analytics
- Hadoop is a powerful analytics tool: No. Hadoop is only for storing and querying data. It is of little practical use at the moment in analytics
- Big data is the next big thing in analytics: No. Most companies have structured data in tables. Big data is applicable only in a few companies
- Data is mostly unstructured: No. A lot of data in organisations is structured, although it exists in silos in various parts of the organisation
- Analytics models require lot of real time computing power: No. Analytics models are built off-line and run with real time data. Running the model requires very little rela time computing power. Building models requires computing power but that is a one time activity and is not real time
Nice thoughts.. Hope to see some good solutions in this space
hey b, thanks for the comment. yes I am planning to start an open source project for implementing a thin warehouse..
Agree
Things that should never be told to the client!!
Eventually they will get to know anyways 🙂