Salta ai contenuti

Checkpoint challenge

Questi contenuti non sono ancora disponibili nella tua lingua.

Original, scenario-based practice. Never modeled on actual exam questions. It’s written in the “What I Would Architect” style: the point is making your judgment visible, not picking a letter. This challenge closes Part 2 of the Data Architect course; my walkthrough comes in Checkpoint 2, so try it first.

Steve Trucking is the last-mile delivery company from Checkpoint 1. Start from your own Checkpoint 1 design: the delivery core, the partner key, the Recipient, and what you kept off core. Now it has to survive five years of growth, three partner integrations, the retirement of RouteDesk, a busy afternoon, and everyone who wants to see the data: partners, crews, Recipients, applications and agents.

Read the Steve Trucking brief again, sections 1–8. Leave section 9 (the late changes) alone; those come at the end of the course.

Everything in the brief’s sections 5 to 8. The ones this checkpoint leans on most:

  • R1, R2. Manifests in, and status back out, in each Retail Partner’s own format and timing: an SFTP file and a nightly file (Keystone), a REST API and webhooks (Redrock), EDI both ways (Brightwave).
  • R4. A manifest revised after routing: coordinators see the change and the affected route immediately.
  • R14, R15. Finance invoices daily from completed work; RouteDesk’s history stays available for claims and analytics after it’s retired.
  • S2. A Retail Partner never sees another partner’s deliveries, pricing, volumes or Recipients.
  • S3. The tracking link shows one delivery, without an account, and stops working after a set time.
  • S4, S5. Recipient details and photos are personal data. Crews see Recipient details only for their assigned stops, only on the day.
  • R8. Recipients can ask “where’s my delivery?” and get an answer, which means an agent.
  • O1. Three times today’s volume within five years.
  • About 1 million deliveries a year, growing 30% a year; about 1.6 line items, eight status changes and four photos per delivery; about 200,000 GPS readings a day; about 180,000 cases a year.
  • RouteDesk holds eight years, about 5 million deliveries, and is being retired.
  • One Salesforce org in the US with Field Service and the partner portal; Snowflake as the warehouse; an integration layer between the partners and Salesforce.
  • Three Retail Partners, nine distribution centers, and more of both coming.

Take your Part 1 design to year five, and secure it. Use what you’ve learned in chapters 7–12:

  1. Size it (chapter 7): with the Number of Zeros, how many Work Orders, line items and status changes at year five, a year and in total if nothing ever leaves? Place each on the LDV ruler. Which objects become large, and when? (Challenge D2.)
  2. Decide what stays (chapters 7 and 8): what lives in the org, what lives off core, and how long a completed delivery stays in the org. Where does it go when it leaves, how is it found again for a claim, and how does its security survive the move? (Challenge D2.)
  3. Find the hot spots (chapters 7 and 11): which parent records, owners or shared records will concentrate work, and what do you change in the model before they do?
  4. Write the contracts (chapter 9): for every important arrow (three manifests in, three status feeds out, status history to the warehouse, the truck’s position, delivery counts to finance), say who starts it, how fresh it must be, the key, and who proves it arrived. (Challenges D3 and D4.)
  5. Retire RouteDesk (chapter 10): what do you migrate, archive or discard, and how do you prove it’s complete? (Challenge D7.)
  6. Design who sees what (this chapter): Retail Partners and their DC managers, coordinators and customer care, crews, Recipients on the tracking page, the partner integrations, and the agents. For each, what may it see, what may it do, and which mechanism enforces it? Count your sharing rules at year five. (S2–S5.)
  7. Classify and protect (this chapter): which Steve Trucking fields are Confidential or Restricted, and what does that change: encryption, masking, retention, what an agent may read?

Write each decision as both options, then your choice: “in my professional opinion…”

My walkthrough is Checkpoint 2: Steve Trucking at Scale, What I Would Architect. Compare your reasoning, not just your boxes. Where you chose differently, can you say what you’d gain and what you’d give up?

Keep this design too. Part 3 asks what changes when the same data has to serve analytics and AI, and chapter 18’s review card checks the whole design at year three, against 70% of every limit.

  • DM-2, DM-4, DM-5 (the model respects sharing; big objects or off core; data skew)
  • LDV-1, LDV-2, LDV-3 (large data volumes, archiving, performance)
  • MIG-1, MIG-2, MIG-3 (migration and proof)
  • SDM-1 (licenses and access), GOV-1 (classification and retention)