If we need to build a a visualisation to show these loads then we’re interested in a number of things:
- how frequent are the loads per client?
- how big is each load? (we might want to spot outliers in size variation e.g. if a client regularly sends 1000 records and suddenly they send 10 records it might be an issue)
- how successful is each load?
- where in each data load are the errors occurring, is an error throughout the data or is it just on the top X rows, say?
- what specifically is the error for a given line to help us pinpoint the actual problem.
The dashboard below outlines how I approach this situation with an example of a dashboard using fake data I generated in Alteryx. The actual real visualisations included information per column in the sheet on the top right, as well as row by row information on the actual values being imported in the third (bottom right) sheet.
Using Action filters to drill into problem areas and investigate issues was a key requirement and it helped speed up the dashboard here which was hundreds of million rows in some instances. The action filters only focused on the necessary data and kept the speed down to just a second or so for querying the data. Aggregating data and providing data drill through is a key exploratory requirement in many Tableau dashboards.
Click below to explore the dashboard and download it from Tableau Public.