↵ select ↓ ↑ navigate esc close

Week 5 - Seeing How the Business Actually Works

Practical Data Modeling ·

Chapter 13, Seeing the Business, printed pages 265–294

I’m sure you’ve been here. You can build a technically impressive data model and still get the business completely wrong. The tables join. The pipelines run. People look at the results and say, “That’s not how we work.” This week, we’ll spend some time on what you need to understand before those tables are useful.

Chapter 13 asks us to follow the actual process, including the exceptions and workarounds. A process crosses teams; a word's meaning can change at each boundary. Often, organizations have MANY boundaries, so a word’s meaning can be all over the place. Reading the official documentation helps (if it exists and is up to date), but so does sitting with the person who has to make the thing work every day. Their workflow or spreadsheet might tell you something the architecture diagram leaves out.

For this week’s hypothetical complication, our online retailer’s support team wants shorter resolution times. Finance wants lower support costs, and Product wants to identify recurring defects. We still have Ticket, Customer, Agent, Product, and TicketProducts, with tickets arriving through email, chat, and phone. But an agent closing a ticket might look like success to one team while the customer is still waiting for a replacement.

Before adding another metric, I’d want to follow one complaint from the first contact through the replacement and any follow-up. Where does support’s responsibility end? What does each team need to know? Which distinctions must the model preserve? For the discussion, let’s explore how we would find those answers without pretending a single status field can settle every disagreement.

Questions for discussion

  1. Where would you begin and end this retailer’s support process, and which handoff would you investigate first?
  2. What evidence would you ask for before accepting each team’s definition of a successfully resolved ticket?
  3. If an agent’s workaround contradicts the documented process, what would you capture in the model and verify with the process owner?

Optional exercise: Spend ten minutes on this: Sketch one ticket’s journey across Support, Product, and Finance. Mark one point where a term changes meaning and write the question you’d ask the person doing that work. A rough sketch is plenty.

The public comments are open to everyone, whether or not you attend live. If you’re behind, bring one question. Please sanitize any examples from work and leave out customer, employee, and company-confidential details.

Live discussion is for paid Practical Data Modeling subscribers. I’m traveling this week, so check the paid subscriber chat on Substack for the confirmed day, time, and Google Meet link. Message me on Substack if you need the calendar or Meet link.

We’ll pause for Thanksgiving the week of November 25 (US holiday), with no assigned reading or live session. Week 6 picks up the week of December 2.

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