Data-Driven Requirements Gathering

 

Early in my career I was fortunate to have a mentor who was very data driven.  He believed that data in its purest form would help describe the processes that you might want to implement in a system, but that process analysis alone would not properly define the data and in fact, might very well define it improperly due to the usual insular aspect of looking at some processes and not all, of necessity given scope and budgets.

When I think about this at a high level in terms of typical data entities and processes, I have to believe that the metrics would support such a conclusion.  If we look at any particular organization, the number of processes operating on the same, similar or related data will be very high, yet the number of entity types the organization deals with will be very low in comparison.

So let's look at a very simple example, a basic course registration system.  Here are the entity types we'll deal with:

1. Student
2. Course

Here are some business rules around the data:

• A student can register for up to eight courses.
• A course can have up to one hundred students.
• Students can see who else has registered for a course

From this simple example, I conclude that we have at least the following potential processes:

1. Registering for a course
2. Seeing details like who is teaching a course, costs of courses
3. Listing who is also registered for a specific course
4. Updating or deleting a course registration
5. Finding out if I am allowed to delete a course registration
6. Finding out the rules and timelines around course registration
7. Withdrawing from a course after I have already started attending it
8. Listing all courses I as a student am registered for
9. Displaying detailed information about a course I am registered for, or courses for which I am not registered
10. Plotting course schedules on a calendar so I can generate an agenda for myself
11. Show ratings for each course
12. Show ratings for each instructor
13. Show the marks I have achieved for each course by term and final
14. Show a history of my marks in the timeframe I select
15. Show a graph of how my performance has changed over time
16. Show a graph of how instructor performance has changed over time for selected courses or for all courses he/she teaches
17. Update course information - new courses, remove or archive courses, update who is teaching a course
18. Update student information - new students, remove or archive student information, update student demographic, biographical information and other information 

.... and so on ... and so on... and so on...

All of the processes I listed come directly from the two entities, student and course, along with the attributes each has and the relationships they have, one with the other.  I am sure with a little extra thought, the list could be doubled.  I am also sure that some of the processes listed have nothing to do with a student registration system, and more to do with other systems, and so could serve to put some rope around both the data (and its attributes) and the processes to be able to plan multiple projects as part of a program.

Furthermore, since we all know there is a many-to-many relationship between Courses and Students, we are probably missing an Entity Type - one that is fairly obvious anyway - something called "Course Registrations" which becomes a one-to-many relationship with each "Kernel" entity type, resolving the many-to-many relationship by adding the "Associative" entity type.  And we won't even get into the difference between registrations and transcripts, whether things like marks even belong in this model, etc.

I obviously invented  a lot of the processes I listed, and they would likely be very different if I actually had stakeholders to help me along (or would they?).  It may reveal new business rules, such as those student registration privacy or restrictions on instructor performance data, or the absence of processes, such as being able to label other students who are your friends, and seeing which courses they have registered for. 

A very simple example, of course, but from two entities (which eventually became three), I was able to derive at least 18 processes (some actually have multiple processes per line), and I did that as quickly as I could type.  Confirmation with the stakeholder would take much longer, of course.

Trying to create a long list of processes and then deriving the data from them, rather than letting the data drive the derivation of processes, is, I feel, much more complex, time consuming and subject to errors and omissions.

What do you think?  Are you process driven, or data driven?  Were you one and became another at some point in time? 

Before you answer, remember the CRUD (or CUDDL) I mentioned in a previous post.  Don't know what that is?  Look up "Of CRUDs and CUDDLs"  :)