Cohort analysis patterns in Tableau - Dashboard Guide

Dashboard Mode - Cohort analysis patterns

Business Question

How do different customer groups (cohorts) behave over time after their first purchase?

Sample Data

Customer ID	First Purchase Month	Purchase Month	Sales Amount
101	2023-01	2023-01	100
101	2023-01	2023-02	50
102	2023-01	2023-03	70
103	2023-02	2023-02	200
103	2023-02	2023-03	100
104	2023-03	2023-03	150
105	2023-03	2023-04	80

Dashboard Components

• KPI Card: Total Customers in Cohort (e.g., Jan 2023 cohort = 2 customers)
• KPI Card: Total Sales for Selected Cohort (e.g., Jan 2023 cohort total sales = 220)
• Line Chart: Sales Over Months Since First Purchase by Cohort
  Formula: Calculate months since first purchase = DATEDIFF('month', DATE([First Purchase Month] + '-01'), DATE([Purchase Month] + '-01'))
  Plot sales amount by this month offset for each cohort
• Heatmap: Retention Rate by Cohort and Months Since First Purchase
  Formula: Retention = Number of customers with purchase in month N / Number of customers in cohort
• Table: Raw Data Table showing Customer ID, First Purchase Month, Purchase Month, Sales Amount

Dashboard Layout

+----------------------+-----------------------+
| Total Customers KPI   | Total Sales KPI       |
| (Top Left)            | (Top Right)           |
+----------------------+-----------------------+
|                      Line Chart: Sales Over Months          |
|                      Since First Purchase by Cohort        |
+-------------------------------------------------------------+
|                      Heatmap: Retention Rate                |
+-------------------------------------------------------------+
|                      Raw Data Table                          |
+-------------------------------------------------------------+
    

Interactivity

Cohort Filter: Select a cohort month (e.g., Jan 2023) to update all components to show data only for that cohort.
Month Offset Slider: Filter the line chart and heatmap to show data for a specific range of months since first purchase.
Hover on Heatmap: Show exact retention rate percentage for that cohort and month.
Click on Table Row: Highlight corresponding cohort in charts.

Self Check

If you add a filter to select the cohort "2023-02", which components update and what changes do you expect?

• The Total Customers KPI updates to show 1 (customer 103).
• The Total Sales KPI updates to show 300 (200 + 100 sales).
• The Line Chart updates to show sales over months since first purchase only for the Feb 2023 cohort.
• The Heatmap updates to show retention rates for the Feb 2023 cohort only.
• The Raw Data Table filters to show only rows with First Purchase Month = 2023-02.

Key Result

Dashboard showing customer cohorts' sales and retention patterns over months since first purchase.

Practice

1. What is the main purpose of cohort analysis in Tableau?

easy

A. To create pie charts for sales distribution

B. To filter data by geographic location only

C. To group users by their start time and track their behavior over time

D. To calculate total revenue without time context

Solution

1. Step 1: Understand cohort analysis concept

   Cohort analysis groups users based on when they started using a product or service.

2. Step 2: Identify its purpose in Tableau

   It tracks user behavior or retention over time, not just static metrics like revenue or location.

3. Final Answer:

   To group users by their start time and track their behavior over time -> Option C

4. Quick Check:

   Cohort analysis = group by start time and track behavior [OK]

Hint: Remember: Cohorts track groups by start time over periods [OK]

Common Mistakes:

• Confusing cohort analysis with simple filtering
• Thinking cohort analysis is only about total sales
• Ignoring the time dimension in cohort grouping

2. Which of the following calculated fields correctly defines a cohort start month in Tableau?

easy

A. DATEDIFF('day', [User Signup Date], TODAY())

B. DATEPART('year', [User Signup Date]) + 1

C. SUM([User Signup Date])

D. DATETRUNC('month', [User Signup Date])

Solution

1. Step 1: Understand cohort start date calculation

   Cohort start is usually the first day of the period, here month, so DATETRUNC('month', date) is correct.

2. Step 2: Evaluate each option

   Calculating days to today gives elapsed time, not the start month; adding 1 to the year part shifts cohorts incorrectly; summing dates is invalid.

3. Final Answer:

   DATETRUNC('month', [User Signup Date]) -> Option D

4. Quick Check:

   Start month = DATETRUNC('month', date) [OK]

Hint: Use DATETRUNC to get cohort period start date [OK]

Common Mistakes:

• Using DATEPART instead of DATETRUNC for cohort start
• Calculating date differences instead of truncating
• Applying aggregation functions on dates incorrectly

3. Given the cohort start month calculated as DATETRUNC('month', [Signup Date]) and the current month as DATETRUNC('month', TODAY()), what does this calculation return?
DATEDIFF('month', DATETRUNC('month', [Signup Date]), DATETRUNC('month', TODAY()))

medium

A. The number of days since the user signed up

B. The number of months since the user signed up

C. The user's signup date truncated to the day

D. The total count of users signed up this month

Solution

1. Step 1: Analyze the DATEDIFF function

   DATEDIFF('month', start, end) returns the number of whole months between two dates.

2. Step 2: Apply to given dates

   It calculates months between the user's signup month and the current month, showing how many months have passed.

3. Final Answer:

   The number of months since the user signed up -> Option B

4. Quick Check:

   DATEDIFF('month', signup, today) = months since signup [OK]

Hint: DATEDIFF with 'month' counts months between dates [OK]

Common Mistakes:

• Confusing months with days in DATEDIFF
• Thinking it returns a date instead of a number
• Assuming it counts users instead of time difference

4. You created a calculated field for cohort period as:
DATEDIFF('month', [Cohort Start], [Order Date])
but the results show negative values. What is the most likely cause?

medium

A. [Order Date] is earlier than [Cohort Start], causing negative differences

B. The DATEDIFF function does not support 'month' as an interval

C. The calculation should use DATEADD instead of DATEDIFF

D. [Cohort Start] is not a date field but a string

Solution

1. Step 1: Understand DATEDIFF behavior

   DATEDIFF returns negative values if the first date is after the second date.

2. Step 2: Check date order in calculation

   If [Order Date] is before [Cohort Start], the difference is negative, which explains the issue.

3. Final Answer:

   [Order Date] is earlier than [Cohort Start], causing negative differences -> Option A

4. Quick Check:

   Negative DATEDIFF means first date > second date [OK]

Hint: Check date order: earlier date first to avoid negatives [OK]

Common Mistakes:

• Assuming DATEDIFF can't use 'month' interval
• Confusing DATEDIFF with DATEADD function
• Not verifying data types of date fields

5. You want to create a heatmap in Tableau showing user retention by cohort month and months since signup. Which combination of fields and visualization best achieves this?

hard

A. Rows: Cohort Month (DATETRUNC), Columns: Months Since Signup (DATEDIFF), Color: Count of Users

B. Rows: User ID, Columns: Signup Date, Color: Total Sales

C. Rows: Order Date, Columns: Product Category, Color: Average Price

D. Rows: Months Since Signup, Columns: Total Revenue, Color: Cohort Month

Solution

1. Step 1: Identify correct cohort and period fields

   Cohort Month groups users by signup month; Months Since Signup tracks time elapsed.

2. Step 2: Choose visualization layout

   Heatmap uses rows and columns for cohort and period, color shows user counts for retention.

3. Final Answer:

   Rows: Cohort Month (DATETRUNC), Columns: Months Since Signup (DATEDIFF), Color: Count of Users -> Option A

4. Quick Check:

   Heatmap = cohort by period with user count color [OK]

Hint: Heatmap axes: cohort start and months since signup, color by users [OK]

Common Mistakes:

• Using unrelated fields like product category or revenue
• Placing total revenue on columns instead of cohort period
• Not using count of users for color intensity