Recipes for Probably

This section provides step-by-step guides for common analysis scenarios using Probably. These recipes demonstrate how to combine different features of Probably to gain meaningful insights from your data.

Recipe 1: Analyzing Customer Feedback

Scenario

You have a dataset of customer feedback for various products, including the feedback text, product category, and customer satisfaction rating.

Steps

1. Load your feedback dataset into Probably.

2. Start with an overview:

   • Plot "Product Category" (X) vs "Satisfaction Rating" (Y)
   • This gives you a quick view of how satisfaction varies across product categories

3. Dive into the text data:

   • Select "Feedback Text" as your X-axis to enter Cluster View
   • Color-code points by "Satisfaction Rating" (set as Z-axis)
   • Identify clusters of similar feedback and how they relate to satisfaction

4. Explore specific categories:

   • In the X vs Y plot, click on an interesting product category
   • Select "Plot with new X" and choose "Feedback Text"
   • This shows you the cluster view for just that product category

5. Summarize insights:

   • Use the "Summarize All" feature in Cluster View to get an overview of main themes
   • Compare summaries across different satisfaction levels or product categories

Outcome

You'll have a comprehensive view of customer feedback, including overall satisfaction trends and specific themes in the feedback text, segmented by product category.

Recipe 2: Identifying Trends in Time Series Data

Scenario

You have sales data over time, including date, product, sales amount, and region.

Steps

1. Load your sales dataset into Probably.

2. Create a time series overview:

   • Plot "Date" (X) vs "Sales Amount" (Y)
   • Use "Product" as your Z-axis to color-code different product lines

3. Identify seasonal patterns:

   • In the "Advanced Controls", set "Time Grouping" to "Month"
   • This aggregates your data by month, potentially revealing seasonal trends

4. Compare regions:

   • Use "Region" as a confounder variable
   • This creates separate plots for each region, allowing you to compare trends

5. Drill down into specific products:

   • Click on a product line in the legend
   • Select "Apply as filter"
   • Now you're looking at the time series for just that product

6. Analyze year-over-year growth:

   • In "Advanced Controls", select "Year-over-Year Comparison"
   • This overlays data from different years, making it easy to compare growth

Outcome

You'll have a clear picture of sales trends over time, including seasonal patterns, regional differences, and product-specific performance.

Recipe 3: Cohort Analysis for User Retention

Scenario

You have user activity data, including user ID, registration date, and dates of subsequent activities.

Steps

1. Load your user activity dataset into Probably.

2. Prepare your data:

   • Use Probably's data transformation features to create a "Cohort" column based on registration date (e.g., month of registration)
   • Create a "Months Since Registration" column

3. Create a cohort retention plot:

   • Plot "Months Since Registration" (X) vs "Cohort" (Y)
   • Set the Z-axis to count of active users
   • In "Plot Type", select "Heatmap"

4. Analyze retention patterns:

   • Darker colors in the heatmap indicate higher retention
   • Look for patterns across cohorts and over time

5. Drill down into specific cohorts:

   • Click on a cohort of interest
   • Select "Plot with new X" and choose a relevant activity metric

6. Compare cohorts:

   • Use the "Cohort" as a confounder variable when looking at other metrics
   • This allows you to see how different cohorts behave across various dimensions

Outcome

You'll have a visual representation of user retention over time, allowing you to identify successful cohorts and potential areas for improvement in your user retention strategies.

Remember, these recipes are just starting points. Feel free to adapt them to your specific data and analysis needs. The power of Probably lies in its flexibility and your creativity in exploring your data!