Why Good Data Analyst Candidates Still Get Rejected

Job description line            ->  What the team really does

'Build and maintain Tableau       ->  Daily Tableau work; expect a
 dashboards for operations'          live Tableau question

'Work with supply chain           ->  Learn fill rate, stock-out,
 stakeholders'                       lead time before the interview

'Strong SQL on large datasets'    ->  Expect window functions and
                                     performance-aware queries

'Ad hoc analysis under tight      ->  They value speed and clear
 deadlines'                          short summaries

Interviewer: You have mostly used Power BI. We use Tableau.

Candidate: That is right. The core ideas carry over:
a clean data model, calculated fields instead of DAX
measures, and filters and parameters for interactivity.

To check, I rebuilt my sales dashboard in Tableau Public
last weekend. The main things I had to relearn were
level-of-detail expressions, for example
  { FIXED [Region] : SUM([Sales]) }
which is close to what I did with CALCULATE and ALL
in Power BI. I can show you that version.

## Summary (read this first)

Question: Why did customer complaints rise in Q3?

Answer: Most of the rise comes from late deliveries in
two warehouses after a carrier change in July.

Recommendation: Review the new carrier's SLA for those
warehouses; complaints in other warehouses were flat.

Assumptions:
- A complaint with any 'delay' tag counts as delivery-related
- Orders cancelled before dispatch are excluded

Limitations:
- No data on carrier costs, so I cannot judge trade-offs

import pandas as pd

def load_data(path: str) -> pd.DataFrame:
    df = pd.read_csv(path, parse_dates=['order_date', 'delivered_at'])
    return df[df['status'] != 'cancelled_before_dispatch']

def add_delay_flag(df: pd.DataFrame, sla_days: int = 3) -> pd.DataFrame:
    days = (df['delivered_at'] - df['order_date']).dt.days
    return df.assign(late=days > sla_days)

def complaints_by_warehouse(df: pd.DataFrame) -> pd.DataFrame:
    return (df.groupby(['warehouse', df['order_date'].dt.quarter])
              .agg(orders=('order_id', 'nunique'),
                   late_share=('late', 'mean'),
                   complaints=('complaint_flag', 'sum'))
              .round(3))

if __name__ == '__main__':
    orders = add_delay_flag(load_data('orders.csv'))
    print(complaints_by_warehouse(orders))

Subject: Thank you - Data Analyst interview

Hi [Name],

Thank you for the conversation today. I kept thinking
about the question on rising return rates. One thing I
would check first is whether returns are concentrated in
first-time customers, since that would point to listing
accuracy rather than product quality.

I am excited about the role and happy to share more of my
work if useful.

Regards,
[Your name]

AFTER A REJECTION, ASK YOURSELF

Controllable:
[ ] Did I match the team's main tool and tasks?
[ ] Did my take-home lead with the answer?
[ ] Were my project facts consistent across rounds?
[ ] Did I show interest in THIS team's problems?
[ ] Did I ask for feedback politely?

Not controllable:
[ ] Internal candidate / role paused / budget

Fix one controllable item before the next interview.

TEAM'S DAILY WORK (from JD and earlier rounds)

Main tool: __________   My evidence: __________
Main data domain: ____  Metrics I studied: _____
Typical task: ________  Similar thing I did: ___

Gap I must address honestly:
______________________________________
How I will show I can close it:
______________________________________

BEFORE SUBMITTING A TAKE-HOME

[ ] Summary with question, answer, recommendation on top
[ ] Assumptions listed
[ ] Limitations listed
[ ] At most 3-5 charts, each titled with a finding
[ ] Code split into small, named functions
[ ] Runs top to bottom on a fresh kernel
[ ] File and column names are readable

WHY THIS TEAM (not why data)

One problem the team likely faces:
______________________________________
One metric that matters to them:
______________________________________
One question I will ask about it:
______________________________________
One line connecting my project to it:
______________________________________