buildinterview
Sign inPricing
Back to overview

Purpose-built Audition walkthrough

Checkout margin incident

A fictional Ecommerce company sees mobile net revenue fall after a checkout release. The candidate must decide whether to hold, roll back, or continue while the available evidence contains a real measurement limitation.

Target role
Senior Product Analyst
Format
Plan · Build · Writeup
Build work
SQL · Python · validation
Expected artifact
Decision memo

Demo-only product content. This walkthrough has no qid, is not question-bank inventory, does not call a grader or sandbox, and contains no hidden answer, private check, or candidate evidence.

01 · Business brief

Begin with a decision worth making.

The demo is written as one coherent work sample. SQL and Python are tools inside the investigation; they are not disconnected puzzles.

Business Problem.md candidate visible
Northstar Commerce · fictional

Mobile net revenue fell after checkout release 7.4.

Growth wants a fast rollback. Finance believes refunds and discount treatment explain most of the alert. Product suspects the mobile completion event is missing for part of the traffic. Decide what the company should do next.

Reconcile the metric before diagnosing it Separate mix shift from within-segment change State what the available data cannot establish

02 · Schema and Build tasks

See the data model when the work calls for it.

The full schema belongs inside the product exploration, after the buyer understands the Audition. Select either Build task to trace the tables and relationships it uses.

Checkout margin AuditionPurpose-built public walkthrough · not bank inventory
Tables
9
Relationships
12
Build tasks
2
Demo contract mapped
01 · Inspect the demo data

The schema gives the work its context.

Active question path
Selected tableorders1.8m rows3 visible fieldsUsed by the selected work sample
02 · Trace the Build work

SQL and Python use the same world.

SQL40 min suggested workload

Business decision

Did the mobile release reduce true net revenue, or did reporting definitions create the alert?

Bound data

customerssessionsorderspaymentsrefunds

Candidate deliverable

A grain-safe reconciliation with explicit revenue, refund, and device definitions

The tasks belong to one Audition. The candidate must reconcile the same business decision across SQL, Python, and the final memo. No private checks or answers are exposed.

Switch tasks to trace the exact tables used by each part of Build.Demo-only content · no qid · no private question or answer material

03 · Candidate workspace

Files, runtime, outputs, and AI in one room.

This interactive reconstruction demonstrates the Build phase. Review the assistant patch, switch files and panels, then run the analysis to see the saved output.

Executable project Schema-aware context Attributable actions
Interactive reconstruction · Build phaseNo model or sandbox call
ExitCheckout margin incident
Question 1 / 3 46:12 Pause
Plan Build Writeup
34 min elapsed Continue to Writeup
  1. import pandas as pd
  3. sessions = pd.read_csv("data/sessions.csv",
  4. parse_dates=["started_at"])
  5. orders = pd.read_csv("data/orders.csv",
  6. parse_dates=["created_at"])
  8. # inspect the change before writing findings.json
  10. print(f"Loaded {len(sessions)} sessions / {len(orders)} orders")
main Illustrative workspace0.4%PYanalysis.py local

04 · Evidence review

The result stays connected to the work.

The illustrative finding below is not a score or answer key. It demonstrates how a reviewer moves from an evaluation statement to the submitted sources behind it.

Illustrative findingEvidence linked

The candidate challenged the measurement contract before recommending a rollback.

They reconciled orders, payments, refunds, and device traffic, then limited the recommendation because the completion event was not reliable after the release.

SourceInspectable evidenceState

PlanTelemetry risk named before BuildLocked

SQLNet revenue reconciliation by deviceRun 04

PythonMix-shift decomposition and sensitivityv7

WriteupHold, instrument, rerun, then decideSubmitted

Map it to your role

Use the walkthrough to start the real interview design.

Design an interview Talk through your role