How Tableau Helps You Land a Data Scientist Job Faster

import numpy as np
import pandas as pd
from sklearn.linear_model import LogisticRegression

rng = np.random.default_rng(7)
df = pd.DataFrame({
    "customer_id": [f"C{i:04d}" for i in range(400)],
    "tenure_months": rng.integers(1, 48, 400),
    "plan": rng.choice(["Basic", "Standard", "Premium"], 400),
})
df["churned"] = (df["tenure_months"] < 8).astype(int)   # toy label

X = pd.get_dummies(df[["tenure_months", "plan"]], drop_first=True)
model = LogisticRegression(max_iter=1000).fit(X, df["churned"])
df["churn_probability"] = model.predict_proba(X)[:, 1].round(3)
df["risk_band"] = pd.cut(df["churn_probability"], [0, 0.3, 0.6, 1.0],
                         labels=["Low", "Medium", "High"], include_lowest=True)
df["tenure_band"] = pd.cut(df["tenure_months"], [0, 6, 12, 24, 48],
                           labels=["0-6", "7-12", "13-24", "25-48"])
df.drop(columns="churned").to_csv("churn_scores_for_tableau.csv", index=False)
print(df["risk_band"].value_counts())

File: churn_scores_for_tableau.csv

customer_id        Unique customer ID
tenure_months      Months since signup
plan               Basic / Standard / Premium
churn_probability  Model's estimated probability of leaving (0-1)
risk_band          Low < 0.3 <= Medium < 0.6 <= High
tenure_band        Grouped tenure for filtering

Model: logistic regression, version 1.0, scored on (date)

// Tableau calculated fields for the churn dashboard

// High-risk count
SUM(IF [risk_band] = "High" THEN 1 ELSE 0 END)

// Average churn probability (as %)
AVG([churn_probability]) * 100

// Share of customers in High band
SUM(IF [risk_band] = "High" THEN 1 ELSE 0 END) / COUNT([customer_id])

import numpy as np
import pandas as pd

# Summarise an A/B test with a 95% interval for each version (simulated data)
rng = np.random.default_rng(3)
visits = pd.DataFrame({"version": ["A"] * 2000 + ["B"] * 2000,
                       "converted": np.r_[rng.random(2000) < 0.10, rng.random(2000) < 0.12]})
summary = visits.groupby("version")["converted"].agg(["mean", "count"])
se = np.sqrt(summary["mean"] * (1 - summary["mean"]) / summary["count"])
summary["low"] = (summary["mean"] - 1.96 * se).round(4)
summary["high"] = (summary["mean"] + 1.96 * se).round(4)
summary.to_csv("ab_summary_for_tableau.csv")
print(summary)

import numpy as np
import pandas as pd
from sklearn.metrics import precision_score, recall_score

rng = np.random.default_rng(5)
log = pd.DataFrame({
    "week": np.repeat(pd.date_range("2026-01-05", periods=8, freq="W-MON"), 250),
    "predicted": rng.random(2000) < 0.3,
    "actual": rng.random(2000) < 0.3,
})
rows = []
for week, g in log.groupby("week"):
    rows.append({"week": week.date(), "volume": len(g),
                 "precision": round(precision_score(g["actual"], g["predicted"]), 3),
                 "recall": round(recall_score(g["actual"], g["predicted"]), 3)})
weekly = pd.DataFrame(rows)
weekly.to_csv("model_monitoring_for_tableau.csv", index=False)
print(weekly)

Model monitoring dashboard (one screen):

Top row:    Predictions this week | Avg score | Latest precision | Latest recall
Middle:     Line chart of precision and recall by week
Bottom:     Bar chart of prediction volume by week
Filters:    Model version, segment
Note box:   'Accuracy uses outcomes known after 30 days'

Project: Churn risk model with a stakeholder dashboard

- Python: clean data, train and evaluate a model, export scores
- Data dictionary for the export
- Tableau dashboard: risk by segment, action list, KPI tiles
- Experiment page with intervals (simulated A/B test)
- Monitoring page with weekly metrics
- Short story: three slides or captions explaining findings
- README linking the notebook and the published dashboard

1. Question: which customers are likely to leave, and where?
2. Finding: risk is highest for short-tenure Basic plan customers
3. Evidence: segment chart and model evaluation summary
4. Recommendation: test an onboarding offer for that group
5. Caveat: model trained on past behaviour; monitor weekly

Situation: model scores were only in a notebook
Task: make results usable by a business team (practice project)
Action:
  - exported tidy scores with a data dictionary
  - Tableau dashboard with segment and action views
  - experiment and monitoring pages
Result: a clear dashboard I can walk through in five minutes