Tableau or Pandas First: What Beginners Should Learn

# Pandas — clean and prepare data.
# This is the step that makes Tableau (or any viz tool) work.

import pandas as pd

# Load raw data
df = pd.read_csv("sales_export.csv")

# Inspect
print(df.shape)
print(df.dtypes)
print(df.isna().sum())

# Clean — fix types, handle missing, remove duplicates
df["order_date"] = pd.to_datetime(df["order_date"], errors="coerce")
df["amount"] = pd.to_numeric(df["amount"], errors="coerce")
df = df.dropna(subset=["order_date", "amount"])
df = df.drop_duplicates(subset=["order_id"])

# Standardise category labels
df["region"] = (
    df["region"]
    .str.strip()
    .str.title()
    .replace({"Ncr": "NCR", "Delhi Ncr": "NCR"})
)

# Create a derived column Tableau will use
df["month"] = df["order_date"].dt.to_period("M").astype(str)

# Save a clean file for Tableau
df.to_csv("clean_sales.csv", index=False)

# What this does:
# - Converts strings to real dates and numbers
# - Removes rows that would break charts
# - Normalises inconsistent labels (Title Case, merged "Ncr")
# - Adds a month column for time-series charts in Tableau
#
# What this proves in an interview:
# - You know that "clean data" is the pre-requisite for
#   every dashboard
# - You handle the messy real-world cases (dates, types, labels)
# - Your pipeline is reproducible, not re-typed every week

# Pandas — answer a business question.
# This is what you do BEFORE Tableau — the analysis itself.

import pandas as pd

df = pd.read_csv("clean_sales.csv")

# Business question: which region had the highest growth
# from the first half of the year to the second half?

# Step 1: split into halves
h1 = df[df["month"].isin(["2026-01", "2026-02", "2026-03", "2026-04", "2026-05", "2026-06"])]
h2 = df[df["month"].isin(["2026-07", "2026-08", "2026-09", "2026-10", "2026-11", "2026-12"])]

# Step 2: sum revenue per region in each half
rev_h1 = h1.groupby("region")["amount"].sum().rename("revenue_h1")
rev_h2 = h2.groupby("region")["amount"].sum().rename("revenue_h2")

# Step 3: join and compute growth
summary = pd.concat([rev_h1, rev_h2], axis=1).fillna(0)
summary["growth_pct"] = (
    (summary["revenue_h2"] - summary["revenue_h1"])
    / summary["revenue_h1"].replace(0, pd.NA)
    * 100
).round(1)

# Step 4: sort and view
summary = summary.sort_values("growth_pct", ascending=False)
print(summary)

# Output:
#           revenue_h1  revenue_h2  growth_pct
# NCR           850000     1120000        31.8
# West          620000      720000         16.1
# South         480000      490000          2.1
# East          350000      280000        -20.0

# What this does:
# - Computes a real business metric (growth by region)
# - Compares two periods with a clean join
# - Handles missing/zero revenue safely
#
# This is the actual analysis. Tableau will visualise
# "growth_pct" as a bar chart — but the number came from here.
#
# Interviewers ask: "How would you compute growth by region?"
# They want the Pandas answer, not the Tableau one.

-- Tableau — what it actually builds (from the clean data)
-- Under the hood, Tableau sends queries like this to the data source.

-- The "growth by region" bar chart in Tableau runs this query:

SELECT
  region,
  SUM(CASE WHEN month BETWEEN '2026-01' AND '2026-06'
           THEN amount ELSE 0 END) AS revenue_h1,
  SUM(CASE WHEN month BETWEEN '2026-07' AND '2026-12'
           THEN amount ELSE 0 END) AS revenue_h2,
  ROUND(
    (SUM(CASE WHEN month BETWEEN '2026-07' AND '2026-12' THEN amount ELSE 0 END)
     - SUM(CASE WHEN month BETWEEN '2026-01' AND '2026-06' THEN amount ELSE 0 END))
    * 100.0
    / NULLIF(SUM(CASE WHEN month BETWEEN '2026-01' AND '2026-06' THEN amount ELSE 0 END), 0)
  , 1) AS growth_pct
FROM clean_sales
GROUP BY region
ORDER BY growth_pct DESC;

-- What Tableau adds on top of this query:
-- - Interactive filtering (user clicks a region to drill down)
-- - Chart rendering (bar chart with color, tooltip, legend)
-- - Dashboard layout (multiple charts, KPIs, text panels)
-- - Refresh schedules (daily, hourly, live)
-- - Sharing (stakeholders view without coding)
--
-- What Tableau does NOT do:
-- - Clean your data (Pandas or SQL does that)
-- - Answer the business question (you decide what to measure)
-- - Replace analysis — it displays it
--
-- This is why Pandas comes first. The analysis is in the code.
-- Tableau is the presentation layer.

# Interview reality check:
# "Show me a dashboard you built" — great.
# "Walk me through a metric you computed" — this is where
# the Pandas part matters. Tableau dashboards look impressive;
# the numbers behind them are what get you hired.