Marketing Relationship Analysis

Overview

The Marketing Relationship Analysis feature helps users visualize the relationship between a response variable (e.g., pipeline) and a marketing variable (e.g., LinkedIn spend) over a specific time period and granularity (e.g., weekly, monthly, quarterly). The widget generates an XY scatter plot with best-fit curves based on various models and highlights the curve with the best R-squared value.

Key Features

1. Visual Representation:

   • Generates an XY scatter plot to illustrate the relationship between the selected variables.

   • Includes multiple curve-fitting options for better data analysis.

2. Curve-Fitting Models:

   • Linear: Straight-line relationship for simple proportionality.

   • Logarithmic: Captures diminishing returns without a defined cap.

   • Power Function: Models diminishing returns using fractional power.

   • Hill Function: An S-shaped curve capturing gradual or steep saturation points.

   • Sigmoid: Models S-shaped growth with clear upper and lower limits.

   • Exponential Decay: Captures rapid early growth and clear levelling-off behaviour.

   • Adstock Saturation: Combines delayed and saturation effects for time-dependent variables.

3. Performance Metric:

   • Highlights the best-fitting curve based on the R-squared value to determine the most accurate model.

How to Use

Create a Configuration

1. Enter a name for the config (Spend vs Pipeline weekly Analysis)

2. Select any filters to apply. (Filter out Pipeline through Outbound channels or Spend through offline channels)

3. Select the Time Period for analysis (e.g., Jan 2024 - Dec 2024).

4. Set the Time Granularity:

   • Weekly

   • Monthly

   • Quarterly

5. Choose the Response Variable (e.g., pipeline-generated).

6. Select the Marketing Variable (e.g., LinkedIn spend).

7. Click Generate. It usually takes up to 10 minutes to generate this config.

8. Once the config gets generated, you will see the Visualization

Analyze Visualization

1. View the scatter plot with the selected variables on the X and Y axes.

2. Choose curve-fitting models to display:

   • Linear

   • Logarithmic

   • Hill Function

   • Power Function

   • Sigmoid

   • Exponential Decay

   • Adstock Saturation

Analyze Best Fit

1. The widget automatically computes the R-squared value for each curve-fitting model.

2. The model with the highest R-squared value is highlighted as the Best Fit.

3. Use the detailed tooltip on the chart to view key metrics for each model to see the R-squared value.

Curve Models Explained

1. Hill Function

• Shape: S-shaped curve with explicit saturation.

• Use Case: Ideal for capturing diminishing returns, with flexibility for steep or gradual saturation.

2. Logarithmic

• Shape: Diminishing returns without an explicit cap.

• Use Case: Suitable for simple models where saturation is not critical.

3. Exponential Decay

• Shape: Rapid early growth followed by clear levelling off.

• Use Case: Common for short-term campaigns with strong initial effects.

4. Sigmoid (Logistic)

• Shape: S-shaped curve with symmetric behaviour.

• Use Case: Useful for scenarios with gradual growth and symmetric saturation.

5. Power Function

• Shape: Models diminishing returns with a fractional power.

• Use Case: Ideal for moderate diminishing returns.

6. Adstock Saturation

• Shape: Models carry over and diminishing effects together.

• Use Case: Captures delayed and saturation effects in time-dependent variables.

Tips for Effective Use

• Select a time granularity that aligns with your analysis goals (e.g., weekly for short-term campaigns, quarterly for long-term trends).

• Use the tooltip insights to better understand the implications of each curve model.

• Focus on the best-fit model to guide strategic decisions.