↵ select ↓ ↑ navigate esc close

Week 4 - Keeping Track of What We Knew and What We Meant

Practical Data Modeling ·

Read Chapters 10–12, printed pages 195–262: Why Time Matters in Data Modeling; The Semantic Foundation: Context, Intent, and Action; Stringing Together the Building Blocks.

A support ticket arrives at 11 a.m. and gets marked medium priority. At 2 p.m., a manager notices that the customer has Premium Support. The ticket should have been high priority from the start. Overwrite the priority and the report looks corrected. You’ve also lost the fact that the team spent three hours working with different information. Hey, it happens a lot…

That’s the support example on printed page 259, and it gives us something concrete to work through this week. Chapter 10 distinguishes when a fact applies in the world from when the system recorded it. Keeping both lets us ask what we believe now and what someone could have known when they acted.

Chapter 11 adds another problem: people can agree on the value in a field while disagreeing about what it means. In the support case, does resolved mean the agent closed the ticket, the customer confirmed satisfaction, or a defined period passed without a reopening? A timestamp alone cannot settle that disagreement. We need a definition that fits the question and that users understand.

Chapter 12 brings the building blocks together. Its synthesis checklist starts with the business question, then works through data forms, shared entities, grain, time, and semantics before choosing an implementation pattern. We now have enough of the pieces to test how one decision affects the others. Correcting priority history, for example, can change which tickets belong in an SLA calculation and how we interpret an agent’s performance.

Questions for discussion

  1. What should the system answer about this ticket’s priority as known at 1 p.m., compared with what we know after the correction?
  2. Which definition of resolved would you choose for measuring customer outcomes, and what evidence would it require?
  3. Where would a change in that definition affect the grain, history, or aggregation of our support model?

Optional exercise: Draw two timelines for the priority correction: when each fact applies and when the system learned it. Add the answer to one historical question beside the sketch. There’s no need to implement a temporal database.

Bring your sketch or an unresolved question to the public comments, even if you haven’t finished the reading. Keep any examples from work anonymous and free of confidential information.

November 11 is the reading-week anchor. I’m traveling in Europe, so the paid subscriber chat on Substack will have the confirmed live day, time, and Google Meet link. Live sessions and recordings are included for paid Practical Data Modeling subscribers. Message me on Substack if you need the calendar or Meet link.

Practical Data Modeling is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.