Forum Discussion
Data Modeling Large Number of Facts
Hello,
Im working with healthcare data that is rolled up by day and unit inside of each individual hospital. It looks something like this:
| Date | Unit | Beds Filled | RNs | Beds Needed | Covid |
| 1-1-2019 | 23524 | 22 | 5 | 1 | 9 |
| 1-1-2019 | 897564 | 15 | 10 | 0 | 1 |
I have two other dimension tables. A date dimension table and a hierarchy table that I use to drill down the fact table with. For example the client can drill down to a particular hospital or unit type with the hierarchy table.
The issue is that I have 3 or 4 data sources each with 20-30 metrics. Im concerned that if I joined all of them together, my fact table would be 100+ columns wide and my data model would take forever to refresh. Should I join them all so that I have a true "star schema" data model or keep them split up in a "constellation" model.
2 Replies
- KNP
Super User
hwr7dd - A star schema is always preferred in Power BI. My advice based on the info you've given would be to join them and then remove all columns that are not used in the reporting. There's always the temptation to keep columns you may need but I would advise to get rid of them until you actually need it.
Unfortunately this is the truncated model. We have well over 300 metrics that we "could" report on that have been left out. The model that needs to be developed is fairly specific to healthcare/IoT as it has an unusual amount of facts and a low amount of dimensions.
