By Ryan Trybus, Consultant
In a recent endeavor, I was given the challenge to work with a team that handled customer satisfaction survey response data for their respective company. Until the point of Bardess’ involvement, their practice for displaying data consisted of various charts contained in massive Excel files. Their hope was that I could take their data presented in these static views and create a dynamic experience using Qlikview.
As with all of our BI tool development efforts here at Bardess, our focus in building a house of business analytics is to start with a strong foundation of data integrity. During the early stages of IT Planning and Development, I worked closely with the database administrators as they replicated existing data sources into a server that would directly feed the Qlikview application.
The most interesting aspect of this process came during the QA testing of their data source. The client’s QA methodology consisted of checking that score calculations on the QV dashboard matched the values found in prior Excel reports. What I found in testing was not only did Qlikview’s total number of survey responses differ from the existing reports, but resultantly the monthly and rolling-month averages between the two reports varied up to 1%.
The source of the discrepancy surfaced when I found that Qlikview had discovered and omitted multiple undesirable records. Not only did the dashboard expose a handful of duplicate entries in the source data, but also flagged survey responses that had valid record details but were missing satisfaction scores. In essence, this was adding extra surveys with scores of 0 into the calculations for monthly and three-month rolling averages! What started off as a sleuthing effort to determine the cause for differing report values end up in realization that survey calculations were being understated! These small errors in value couldn’t have normally been spotted in Excel reports with the averages so consistent from month to month.
It’s very easy for us to overlook issues with our data integrity when the results so perfectly fall in line with prior outcomes. We want to help you see the truth in your data. Contact Bardess for more information on our data management, business intelligence and IT Planning and Development initiatives.
About the Author
Ryan Trybus, a Consultant at Bardess, has employed his background in computer science to help promote data automation in the workplace. His focus is providing clients the ability to see the data they need to, as well as share information across their user base as effectively as possible. He comes to Bardess with experience working with a range of traditional business reporting and asset management tools – from Microsoft’s software suite to more specialized solutions like IBM’s Maximo product. He graduated from Arizona State with a B.S of Computer Science with a focus in Mobile development and UI interfaces.