Hi, our team is fairly new to DV modeling, we have one resource with a few sparse years of experience and I have many years of modeling experience but little DV. We have a source file that provides convenience store information. Some attributes are
store manager name
regional manager name
- many more, possibly 50 in total
in addition to store, we will get sales transaction that reference store (we can discuss this later but do know that we will have other data that references store).
my question is how to model this. from what i have learned, i would create a single hub (store id key) with a store satellite which contains most of these attributes. We will have > 1 address so those would go into another multi-active satellite linked to the store hub (customer address, delivery address, municipal address). As of now, i would not split out any of the other attributes into their own satellites as we don’t have accompanying information to go along with those satellites. i.e. i would not split the regional manager into a separate satellite as all we have is the person’s name.
Now the model that was created splits just about 15 attributes into their own hub without a satellite, and those hubs are related to the store hub thru a link. so we have a very large link table with 16 keys, one key for the store itself, and then keys for each of the hubs. most of these hubs dont have a satellite and the key is just the attribute itself, like manager name, regional manager name, region name, etc.
this doesn’t make sense to me, but the modeler is adamant that this is correct. his design has many hubs without satellites, he calls them “stubbed hubs” and if/when we get additional information for these hubs, he will then create the satellites. thing is, we will likely not even get additional information for these stubbed hubs. Data has not realy changed much in 10+ years.
and his design relates the addresses to the store via a link table vs using a multi-active satellite.
any thoughts, anyone willing to talk about this in detail?