Mining the Data Store

A few days ago I sat through a very interesting session, a statistics class (did I say interesting?). This was one in a long-running series we’ve been going through, helping our staff better understand, analyze, and use data. In this particular session we used work order data on carpet cleaning requests as our example, and learned a few things. But specifically, this session was about how to ask questions of the data. Here is a short outline that you can use to ask your own data questions. Listen. You might be surprised at what it tells you!

Getting your head around the problem:

• State the problem (We have too many carpet cleaning requests to handle without affecting other work)
• Research what question to answer with data (why is carpet cleaning one of the top three work requests?)
• What questions will help you answer the research question (what, who, where, when, why, how much)?

Getting your head around the data:

• What datasets will help answer the questions? (where can you get reliable information from?)
• Which observations from these datasets? (what does the information tell you?)
• What time period should you focus on? (can you clarify the observations based on delimiters, such as time of day, etc.?)
• What are potential confounders (issues that muddy the data or conclusions)?
• Assumptions to check

Getting your head around the solution:

• Possible actions, answers, solutions to the problem

In our case we discovered a disconnect between contracted service levels and staff expectations, that the incident count was too high due to multiple requests for individual incidents, that we have a very small number of staff who produce a disproportionate number of the requests, and that a small number of spaces represent a significant number of requests. All of which led to action plans to increase service in problem areas, to better communicate expectations with staff, and to clean our data before using it for analysis.