Plotting Retention Heatmap

I rewrote the cohort analysis — in Python — using Pandas, Numpy, and the simulated weekly dataset in .csv format. The choice to select Python to continue with the script to simplify the process, and the libraries can help a lot.

Main Goals of the day:

Replicate cohort analysis in Python
Generate Retention Heatmap (Day 0 to Day 6)
Save output as CSV + PNG for Tableau later

Step by Step

📍 Step 1: Merged first_deposit with full session data
📍 Step 2: Calculated day_diff = (session_date - cohort_date).days
📍 Step 3: Filtered to day_diff <= 6
📍 Step 4: Used groupby(['cohort_date', 'day_diff']) → counted unique players
📍 Step 5: Pivot → normalized by cohort size → got retention %
📍 Step 6: Plotted heatmap → saved as intra_week_cohort_heatmap.png

Challenges / Insights

I did it in a few lines of Python, comparing the same script in SQL.
And I understand it better than if I’d just run a query.

The script shows immediately the result:

Players who deposited on Sep 4-7 → 75% average retention till Day 6
Players who deposited on Sep 5–7 → 63% average

Why?
Because Sep 5-7 = weekend.
Sep 4–7 = time to prepare for the weekend.

Just an example for the importance of timing strategy for a marketing campaign.

Code Snippet Final

cohort_retention = cohort_pivot.divide(cohort_size, axis=0).round(3) * 100
sns.heatmap(cohort_retention, annot=True, fmt='.0f', cmap='Blues')
plt.savefig('reports/intra_week_cohort_heatmap.png')

</pre>

Next Step

👉 Tableau report → Time to visualize the results.