Paths Subjects Questions Quizzes Pricing Search
Advanced Open Pro

Real-Time and Lakehouse Revenue Numbers Disagree

An analyst reports that the real-time dashboard showed $482,300 in revenue for a given hour last Tuesday, but the lakehouse-backed batch recompute for the same hour (run the following night, per Step 6's reconciliation) shows $479,650 — a roughly 0.6% discrepancy. They ask whether this means the pipeline has a bug.

  1. Explain why a small discrepancy between the real-time number and the batch-reconciled number is not automatically a bug, referencing the specific design decision that makes this expected behavior in some circumstances.
  2. Describe exactly how you'd investigate to distinguish "expected, designed behavior" from "an actual correctness bug," and what you'd look for in each case.
  3. The analyst's dashboard tool only ever shows the real-time number, even for data from weeks ago, and never automatically shows the reconciled batch number once it's available. Is that a problem, and if so, what would you change?

Share this question

← Back to Case Study: Design a Streaming Event Pipeline practice

We use cookies for product analytics to improve OmniAtlas. See our Privacy Policy.