My first Publication Agile-Data-Warehouse-Design-eBook | Page 287

Dimensional Design Patterns for Cause and Effect
267
Modeling Multi-Valued Groups
Sporting events in the previous example are multi-valued indirect causal factors.
Because they are not fundamental details of product sales events—they are only
related to time—the multi-valued modeling challenge they represent can be ad-
dressed after the sales star is implemented or (better still) ignored altogether. This
is not the case when a direct causal factor is a significant multi-valued detail of an
event. For example, imagine you are modelstorming Pomegranate’s medical
insurance claims. When you ask the stakeholders “Who does what?” they reply:
Multi-valued (MV)
event details are
discovered by telling
group themed
stories
Doctor claims amount.
By working through the 7Ws you discover that a DOCTOR (who) claims an
amount for a TREATMENT (what) given to a PATIENT [Employee] (who) on a
specific TREATMENT DATE (when), as shown in the BEAM ✲ event table, Figure
9-6. These are all single-valued details that convert readily to dimensions. But a
problem arises when you come to the why question:
Why does a doctor treat a patient?
When you ask for examples for the resulting DIAGNOSIS why detail, you discover
that a claim contains multiple diagnosis—there is typically more than one thing
wrong with a patient—and the diagnosis codes (ICD10 codes) submitted as part of
every claim are not linked to the specific treatments.
Figure 9-6
MEDICAL
TREATMENTS
event table
containing
group themed
examples
You capture this business knowledge about multiple diagnoses by marking
DIAGNOSIS as MV to denote a multi-valued detail. Generally you discover MV
details by getting stakeholders to tell group themed stories. You do this after they
have told all their other themed stories (typical, different, missing, repeat) by point-
ing at each detail and asking stakeholders if they can give you an example of the
same type of event that would contain groups of that detail; e.g. a group of custom-
ers or a group of products. For most events and most details they won’t be able to
because multi-valued details (and multi-level (ML) details, which you also find this
way) are the exception rather than the rule—thankfully.
Multi-valued (MV)
and multi-level (ML)
event details are
discovered by telling
group themed
stories