The first rule of improv comedy is to AGREE — to say, “YES AND” and don’t block. I can think of a couple of scenarios where this approach will benefit your business intelligence program.

When loading data into your acquisition layer never stop the flow of data for any reason – say YES AND. Even if there are data integrity issues let the data come in. This layer is supposed to be an exact copy of the source data including problems like bad naming conventions, missing data, and duplicate keys. Don’t worry! Using good data quality practices like assigning quality to each row will allow you to prevent the data from flowing any further. Also, reporting limited insights and notating less credible insights based on poor source quality will drive change to improve the source. This is part of the process.

Another YES AND – never limit metrics from joining a fact table because one or more dimensions cannot be located. It would be terrible to limit a cross-process measurement like Sales Amount because one process had a missing dimension. For example, what if marketing only knew the Referrer half the time and they chose to only measure the Sales Amount in their model when they had that information. This filtering would lead other processes leveraging that model to unknowingly analyze data with an invalid Sales Amount. Instead, always use an inferred dimension – a place holder dimension value that indicates the dimension is unknown or missing, but the fact measurement is still available.

We have a lot of fun with analytics at ConradBI. Partnering with us will not only give your company great insights, but those team members we work with will probably have a blast finding them!