How I Spot a Copy-Pasted Data Analyst Project in Minutes

Interview Insights · Data Analyst Projects

How I Spot a Copy-Pasted Data Analyst Project in Minutes

SECTION 01The signals I notice before the call starts

Before I meet a candidate, I read their resume line and, if a link is given, skim the repository or dashboard. Certain patterns make me plan a deeper project round. None of them proves copying on its own. Together, they tell me where to probe.

The strongest signal is wording. When a resume says 'Performed EDA and visualization to derive insights', I learn nothing about the project. When it says 'Found that late deliveries in two cities explained most one-star reviews', I know there is a real finding to discuss.

Sales Data Analysis | Python, Pandas, Matplotlib
- Performed data cleaning and EDA on sales dataset
- Created visualizations to derive insights
- Used Power BI to build dashboard

(Every word here could describe a thousand projects.)

Superstore Returns Analysis | Python, Pandas, Power BI
- Asked why the Furniture category had the lowest profit
  despite high sales
- Found that deep discounts on Tables drove most losses;
  orders above a 20% discount were mostly loss-making
- Built a one-page Power BI view of profit by discount band
  for a category manager

(Same famous dataset, but a clear question and finding.)

The second version uses the same well-known retail dataset that thousands of learners use. That is not a problem. What changed is that there is a specific question, a specific finding, and an intended user. I can now ask real questions about it.

SECTION 02Three live probes that expose borrowed work

In the project round, I do not ask candidates to recite their project. I ask about decisions. Copied work has outputs but no decisions behind them, so these probes quickly reveal whether the candidate was present when the choices were made.

The probes are not tricks. A candidate who did the work answers them easily, often with enthusiasm, because they remember the struggle. A candidate who followed a video answers with the tutorial's words or with a long pause.

import pandas as pd

df = pd.read_csv('superstore.csv', parse_dates=['Order Date'])

# Original project: monthly profit
monthly = (df.set_index('Order Date')
             .resample('MS')['Profit']
             .sum())

# Interviewer asks: show it by quarter instead.
# An owner knows exactly which line to change.
quarterly = (df.set_index('Order Date')
               .resample('QS')['Profit']
               .sum())

print(quarterly.tail(4))

import pandas as pd

df = pd.read_csv('superstore.csv')

bins = [-0.01, 0.0, 0.1, 0.2, 0.4, 1.0]
labels = ['0%', 'up to 10%', '10-20%', '20-40%', 'over 40%']
df['discount_band'] = pd.cut(df['Discount'], bins=bins, labels=labels)

summary = (df.groupby('discount_band', observed=True)
             .agg(orders=('Order ID', 'nunique'),
                  profit=('Profit', 'sum'),
                  loss_share=('Profit', lambda s: (s < 0).mean()))
             .round(2))
print(summary)

# Owner can explain why the bins were chosen and
# what the loss_share column means.

The change probe is my favourite. Being asked to modify one thing, such as the time grain or a filter, is trivial for someone who wrote the code. For someone who copied it, finding the right line in an unfamiliar notebook under pressure is very difficult.

SECTION 03Turning a tutorial project into work you can defend

Most freshers start with a guided project, and that is a reasonable way to learn. The problem is presenting the tutorial output as your project without adding anything of your own. The fix is not to hide the tutorial; it is to extend it.

A simple rule works well: keep the tutorial as your starting point, then add at least one new question, one extra data-cleaning decision, and one output aimed at a specific user. Those three additions give you material that no other candidate has.

# Profit Leakage in Furniture Sales

## Question
Why does Furniture have high sales but low profit?

## Data
Public retail sample dataset (orders, products, regions).
Started from a guided tutorial; the discount analysis,
return handling and manager view are my own additions.

## Key cleaning decisions
- Negative quantities treated as returns, kept separate
- Duplicate Order ID + Product ID rows removed

## Finding
Orders with discounts above 20% were mostly loss-making.

## Limitation
No supplier cost data, so margin causes are inferred.

"I started this from a guided tutorial to learn pandas.
After finishing it, I asked my own question about why
Furniture made losses, added a discount-band analysis
and separated returns, which the tutorial did not do.

The discount finding and the manager dashboard page
are fully my own work. Happy to walk you through them."

Saying openly that a project started from a tutorial does not hurt you. It actually builds trust, because the interviewer now believes everything else you claim. What hurts is being caught claiming work you cannot explain.

SECTION 04Project ownership checklist before any interview

Run this checklist on every project listed on your resume. If you cannot tick a box, either fix the project or remove it. One defended project is worth more than four borrowed ones.

DECISIONS I CAN EXPLAIN

[ ] Why I chose this dataset and question
[ ] Every column I dropped, and why
[ ] How I handled NULLs, duplicates and outliers
[ ] Why each chart type was chosen
[ ] Why I picked these filters or date ranges
[ ] One thing I tried that did not work

CHANGES I CAN MAKE IN UNDER 2 MINUTES

[ ] Change the time grain (month -> quarter)
[ ] Add or remove a filter (one region only)
[ ] Swap a metric (sum of sales -> average order value)
[ ] Add one new column or calculated field
[ ] Re-sort or re-label a chart

Practise each one with a timer.

MY 90-SECOND PROJECT STORY

Question:   ____________________________
Data:       ____________________________
One messy thing I fixed: _______________
Key finding (with a number from my work): ______
Who would use it and how: ______________
Limitation: ____________________________
What I would do next: __________________

SECTION 05Habits that make your projects obviously yours

These habits leave fingerprints on your work that interviewers recognise immediately.

• Keep a project diary — Write two lines after each work session about what you tried and what broke. It becomes your memory in the interview.
• Title charts with findings — Replace 'Sales by Region' with 'West leads sales but loses money on furniture'. It shows you looked at the output.
• Commit small and often — A history of small commits shows how the project evolved, which a single upload never can.
• Credit your starting point — Mention the tutorial or course you started from. Honesty makes the rest of your claims believable.
• Rebuild one project from scratch — Close the tutorial and rebuild the analysis without looking. Wherever you get stuck is what you need to learn.

SECTION 06Interview Q&A — Copied projects

SECTION 07Test yourself — Copied projects quiz

SECTION 08Frequently asked questions

SECTION 09Related reads