What shape is your data in? Is donor information being entered correctly? Are you missing important details? Are the terms and units of your data consistent? If you’re not sure of the answers to these questions you need to conduct an audit of your data.
A systematic review of your data will help you understand how “clean” your data is and will give you insight into the value of your existing data. An audit can also reveal poor data entry processes and gaps in your data entry and management guidelines.
Here are a few tips for conducting an audit that can help you improve the quality of your data.
- Run Exception Reports: An exception report is any report that targets a particular abnormality in your data. For example, you can run a report that shows every record where the “Age” field is left blank. Or, if you have two datasets, an exception report can bring up records that are inconsistent.Many database systems offer built-in exception reporting capability and often you can automate these reports, but you may need to customize your reports to get the kind of exception report you really need.
- Sample Your Records: Reports are a good way to address known issues, but more often you won’t know about a problem with your data until you find it. Manually going through every record is probably not a good use of your time, but if you sample your records you’re very likely to find any hidden data issues.Sampling should be both systematic and random. You should make sure they every type of data you collect is sampled. For example, if you track data for students, teachers, and parents, then you’ll need to make sure that you’ve sampled a significant portion of each group. The number of samples per type can vary depending on the size of your organization and the data set, but 20 is a good solid sample size for most organizations.Then, within the type of data you’re sampling, make sure that you’re not looking at a similar cluster of records. For example, if you review the data for all the students who live in the same neighborhood, you might only be looking at the work of one person entering data. If that one person is exceptionally good, you will never notice the kinds of mistakes other people are making.
- Ask What’s Missing: This is tricky, but it’s possible to look at your data and determine whether important information is routinely missing. For example, if you run a report that shows cases closed by caseworker you might discover that one person has an extraordinarily low number of closed cases. It’s possible that the staff member is just not working very hard, but if you have a sense that they are doing their job, then it’s very possible that the information is not making it into the system.
- The Goal is “Clean Enough” Data: Your data will never be perfect, but if you can identify systematic issues you can address them and then feel reasonably confident that your data is giving you useful information.