Why Every Data Scientist Should Learn Tableau in 2026

import pandas as pd

df = pd.read_csv("subscriptions.csv", parse_dates=["start_date", "end_date"])
df["tenure_days"] = (df["end_date"].fillna(pd.Timestamp("2026-11-01"))
                     - df["start_date"]).dt.days
df["churned"] = df["end_date"].notna().astype(int)
cols = ["customer_id", "plan", "city", "channel", "tenure_days", "churned", "monthly_fee"]
df[cols].to_csv("tableau_subscriptions.csv", index=False)

Exploration log (keep while using Tableau)
View 1: churn rate by plan -> Basic plan much higher
View 2: churn by channel -> partner channel stands out
View 3: tenure distribution -> spike of churn around day 30
Hypothesis: onboarding issue in first month for partner sign-ups
Next: test in Python with a proper comparison

// Churn Rate
SUM([churned]) / COUNT([customer_id])

// Revenue at Risk (customers predicted to churn)
SUM(IF [churn_score] >= 0.5 THEN [monthly_fee] END)

// Customers per City (LOD)
{ FIXED [city] : COUNTD([customer_id]) }

// Parameter: Risk Threshold (float, 0.1 to 0.9, step 0.1)
// Calculated field using it:
IF [churn_score] >= [Risk Threshold] THEN "Target" ELSE "Ignore" END
// Users move the slider to see how many customers
// would be targeted at each threshold.

Story points outline
1. Churn rose last quarter (context)
2. Most churn happens within the first 30 days (finding)
3. Partner-channel sign-ups churn most in that window (evidence)
4. Model flags at-risk customers in week 2 (method, briefly)
5. Recommendation: onboarding call for flagged partner sign-ups (action)

Weak title:   "Churn by channel"
Strong title: "Partner sign-ups churn twice as often in month one"
Weak title:   "Tenure distribution"
Strong title: "Most cancellations happen before day 30"
Write titles that state the finding.

Hand-off checklist (Python -> Tableau)
[ ] One row per entity (customer, order, test)
[ ] Clear column names, no codes without labels
[ ] Dates as proper dates
[ ] Model scores and version column included
[ ] Sensitive fields removed
[ ] Data dictionary saved next to the extract

Dashboard brief
Audience: retention team
Main question: who should we contact this week?
Key views: risk by plan, risk by channel, target list
Filters: city, plan, threshold parameter
Refresh: weekly after model scoring

5-minute meeting script
0:00  The question and why it matters now
1:00  The main finding (one chart)
2:00  Supporting evidence and how sure we are
3:30  Recommended action and expected effect
4:30  What we need from you today