Untangle Conflicting Metrics and Ship a Trusted Dashboard Before the Board Meets
You are a Forward Deployed Engineer at an operations intelligence software company. You have just been assigned to Vertex Manufacturing, a mid-market industrial manufacturer with \600 employees across two production facilities. Vertex…
Create a free account to upload your work. Your progress saves as a draft until you submit.
What You'll Be Doing
You are a Forward Deployed Engineer at an operations intelligence software company. You have just been assigned to Vertex Manufacturing, a mid-market industrial manufacturer with ~600 employees across two production facilities. Vertex licensed your platform three weeks ago. This is your first onsite.
The VP of Operations, Dana Reyes, has one sentence for you when you arrive: "Our production metrics dashboard takes 20-plus minutes to load and nobody trusts the numbers anymore." Dana has a board presentation in 5 days where she will present production KPIs to the CEO and two board members. She needs the dashboard working and trustworthy before then.
During your first hour on-site, you uncover the following:
What you find during site discovery**
- The dashboard queries Vertex's on-prem SQL Server database directly in real-time, with no caching layer. The SQL Server is also running their ERP — every dashboard query locks production tables during peak shift hours (7–9am, 3–5pm).
- Three core metrics are defined differently between the dashboard and the manual spreadsheets the floor supervisors use: OEE (Overall Equipment Effectiveness), Scrap Rate, and Throughput. The dashboard values and the floor values diverge by 8–23% depending on the metric.
- The IT team is two people. The senior IT lead, Marcus, is on a scheduled vacation and won't return until Day 4. The remaining IT generalist (Jamie) has basic SQL skills and can follow documented procedures but cannot debug application code independently.
- Vertex's infrastructure: on-prem SQL Server (cannot be moved or replaced per IT policy), AWS S3 for data archiving (already in use), BI Reporting Tool (already licensed). No Kafka, no streaming infrastructure.
- Dana does not know about the metric definition discrepancy. She believes the dashboard numbers are accurate — just slow.
Your job is to diagnose the real problem, build a working solution for the highest-priority issue, and brief both Dana and Jamie before you leave the site. You have 60–70 minutes.
How Your Work Will Be Scored
What to Submit
Primary Artifact
Format: .pdf, .doc, .docx, .rtf, .txt, .md
A working, demonstrable solution for the root-cause issue you've identified. This must be code or configuration that runs — not pseudocode, not a spec.
-
Your artifact may be: a Python script, a SQL query optimization, a lightweight caching layer, a data pipeline component, or a metric reconciliation utility — whichever you determine is the right first fix given the constraints.
-
The artifact must be deployable within the 5-day window by someone with Jamie's skill level, or must require Marcus only for the initial setup.
-
Include brief inline comments where the logic is non-obvious.
Sign in to upload files
README
Format: .pdf, .doc, .docx, .rtf, .txt, .md
A structured written assessment with three distinct sections.
-
Section A — Diagnosis & Recommendation: What is the real problem? What did you find beyond the slow dashboard? Which issue do you address first and why? What did you explicitly decide not to fix in this window, and why? What are the risks of your approach?
-
Section B — Stakeholder Communication: Two sub-sections, each written for its target audience:
-
(B1) Executive Summary for Dana Reyes (VP of Operations): A 5-bullet summary — no jargon — covering what you found, what you fixed, what she needs to decide, and what she can tell the board. Dana does not have a technical background.
-
(B2) Technical Handoff Note for Jamie (IT generalist): What you built, how to maintain it, what to watch for in production, and who to call if something breaks. Assume Jamie can follow documented steps but cannot debug application code.
-
Section C — AI Usage Log: We expect you to use AI tools. We evaluate how you use them — not whether you use them. Evidence of iteration, redirection, and critical evaluation scores higher than a polished output with no process documentation. The single highest-signal indicator: your video answer to the mandatory AI question. If you cannot name a specific moment where you redirected AI output, evaluators will assume you did not..
Sign in to upload files
Video Walkthrough
Format: .mp4, .mov, .webm
Record as an MP4 or MOV file and upload directly on the Provn platform.
-
Walk through your diagnosis: what did you find, what did you decide was the real problem, and what did you cut from scope?
-
Demo your working artifact — show that it runs. You may simulate the environment with mock data.
-
Answer the mandatory AI question: "Walk me through one moment during this challenge where you redirected the AI — what did it give you, why wasn't it right, and what did you do differently?"
Sign in to upload files
Create a free account to upload your work. Your progress saves as a draft until you submit.
On this page