A simple KMDS analytics reporting project

The task details for this example are as follows:

The SBA connects small businesses with lenders (for more details see https://www.sba.gov/). The SBA can gaurantee loans to help small businesses with cash flow for operations (7a loans). Every year they publish the performance of loans they gaurantee. This can be obtained from the data.gov website. Suppose you want to understand what kind of businesses are cash strapped and how succesfull these businesses were in paying back loans in the last year. For example, you are a journalist, or an analyst for an investment firm and you are interested in understanding the situation of cash flow loans by industry and state.

This can be generated from the published data by:

• Grouping the loan payback data by the business type and then counting the loans that are paid in full, charged off, or are exempt from such analysis by regulation.
• Grouping the loan payback data by state and then obtaining the counts for each loan payback state

This data is published once a year. The data and the data dictionary are available in the repository. You want to document the process of exploring the data and the details of computing the loan performance (as mentioned above) so that when next year comes around, it is easy to get the details of facts discovered during exploratory analysis and computing the performance. It is then a snap to redo the report with the new data. This is done once a year, so it may be overkill to set up a job for it, but doing so is trivial. See this notebook for the analytics component and this notebook to review observations from the previous year's analysis.

The guidelines for the observation types are used to capture the exploratory and data representation observations. This example does not involve any modeling, so no modeling observations are needed.