Model insights

Each visualization in model insights is designed to help you to:

• Visualize and quantify the impact of your organization’s marketing activities.
• Identify which channels are high-performing.
• Identify which channels might need optimization.

These insights then help you to support resource prioritization and allocation.

To view model insights, in the Models interface in Mix Modeler:

1. From the Models table, select the name of a model that has a Last run status of ● Success.

2. From the context menu, select Model Insights.

You see when the specified model is last refreshed and visualizations are displayed using four tabs: Model insights, Attribution, Factors, Diagnostics, and Historical overview.

You can change the date period on which the visualizations on each of the tabs are based on. Enter a date period or select to select a date period.

Model drift

If model drift is detected on the model, you will see a Model drift detected dialog with options to be reminded later or to immediately Retrain the model. If you select Remind me later, you are reminded the next day or at the next login.

Model insights

The Model insights tab shows visualizations for Contribution by date and base media, Contribution by channel, Marketing performance summary, and Marginal response curves. The tab also provides a Touchpoint breakdown table.

• You can hover over individual chart elements in each visualization to display a popover with more details.

• To download a CSV file containing the data for the visualization, select .

• To download full model insights data in Microsoft® Excel format, select Download data.

Contribution by date and base media.

This stacked graph visualization is ordered as follows:

• Base at the bottom.
• Non-spend channels in the middle.
• Spend channels on top.

This visualization represents the contribution proportion achieved by base, by spend channels, and by non-spend channels, across a date range. This visualization is helpful to showcase incrementality. The base represents what would have happened without any marketing at all, and the non-spend channels plus spend channels (on top of the base) attribute to your marketing’s impact. In short, non-spend plus spend equals the incremental impact of your marketing efforts and the visualization provides easy insight into the value that marketing generates.

Contribution by channel

A donut visualization that shows a distribution of the contribution by various channels. This visualization showcases incrementality through the lens of the top three performing channels (excluding base and All others categories). The visualization helps to support prioritization and budget allocation.

Marketing performance summary.

A horizontal bar graph visualization that displays the ROI or CPA performance by each of the channels. This visualization highlights the ROI / CPA of your marketing investments. The channels are ranked in descending order based on ROI / CPA. The visualization helps to identify which channels are most effective and which might need optimization.

Marginal response curves.

The line chart visualizes and compares the marginal returns generated by the investment in your marketing channels. And identifies the break-even point where your incremental return is less than your incremental spend. As a result, this visualization helps you to understand when your marketing investment starts to become less impactful.

The curve, break-even point, and corresponding values are calculated based on the selected data range and the channel you have selected.

To change the channel:

• Select a channel from the Channel dropdown menu to update the visualization for a specific channel.

Touchpoint breakdown

The touchpoint breakdown table shows the weekly touchpoint breakdowns for all or selected channels on a weekly base, displaying key metrics associated to each. The table allows for easy comparison, trend identification and performance tracking at a more granular channel level. This table explicitly complements the Contribution by date and base media visualization and the Contribution by channel visualization.

The following columns are available:

To select a specific channel or all channels, select from the View dropdown menu.

To download the contents of the Touchpoint breakdown table, select Download CSV.

Factors beta

The Factors beta tab shows external factor related insights.

This visualization helps you to understand the incremental effect that various internal and external factors have on the baseline of conversions. For example, economic conditions or promotional activities.

Use the Factors dropdown menu to select which factors you want to display.

To download a CSV file containing the data for the table, select .

If no data is available you see a message No data is available, you may need to retrain your model, or change the date range to view insights.

Attribution

Using the Attribution tab, you can understand the effectiveness of touchpoints and marketing campaigns that have event level data. See Build model.

The following attribution models are supported:

• Based on the selected model in Mix Modeler:
  • Algorithmic - Influenced
  • Algorithmic - Incremental
• Rule based:
  • Decay units
  • First touch
  • Last touch
  • Linear
  • Ushape

See Multi-touch attribution for an introduction on the multi-touch attribution capability in Mix Modeler.

Select one or more attribution models from the Attribution Model dropdown menu. The selected attribution models apply to all visualizations in the Attribution tab.

The Mix Modeler multi-touch attribution granular event scores align to the overall Mix Modeler scores and ROIs. These scores are also made available as datasets in Experience Platform.

The Attribution tab consists of the following visualizations:

Overview

The Overview visualization shows, for the selected attribution models, the conversions totals and percentages. Selecting more models adds additional circles to the visualization, each with its own color corresponding to the legend.

To see a popup with details for an attribution model, hover over any of the circles in the visualization.

Trends

The Daily trends, Weekly trends, or Monthly trends visualization shows, for the selected attribution models, the daily, weekly, or monthly conversion trends.

To choose the period, select Daily trends, Weekly trends or Monthly trends from .

To see details, hover over the data line of a specific attribution model to display a popover that shows the total number of conversions for that data.

Breakdown

The Breakdown visualization is a breakdown by channel or touchpoint of the conversions for each of the selected attribution models. This visualization can be helpful to make decisions on the effectiveness of each channel or touchpoint.

To choose the breakdown type, select Breakdown by channel or Breakdown by touchpoint from .

To see details, hover over any of the chart elements.

Top campaigns

The Top campaigns visualization shows a table of the top campaigns with columns for Campaign name, Channel, Media type and Incremental conversions. This visualization can help inform your team of the effectiveness of a specific campaign for a given channel and provide insights on what campaigns you should further invest into.

To sort the table in ascending ↑ or descending order ↓ for Channel, Media type or Incremental conversions, select the column header and toggle the sort.

To expand the table in a separate dialog, select Expand from .

The expanded Top campaigns dialog shows the same table with addition columns for

• Incremental conversions

• Influenced conversions

• First touch conversions

• Last touch conversions

  You can select each of the additional column headers to sort the table in ascending or descending order.

To close the expanded Top campaigns dialog, select Close.

Breakdown by touchpoint position

The Breakdown by touchpoint position visualization is a breakdown of attributed conversions by position of the touchpoint and touchpoint across all the conversion paths. This chart helps you to compare if a touchpoint contributes better at a position than remaining positions and other touchpoints at any position.

The positions Starter, Player and Closer are defined as follows:

Top conversion paths

The Top conversion paths visualization shows the top 5 conversion paths based on the selected attribution models.

For each conversion path, you see:

• the number of channels that do have an impact,
• the total attributed paths,
• the percentage of attributed paths for this conversion path vs. total attributed paths,
• for each channel, the attribution model contribution percentage, and
• the sum of these channel attribution model contribution percentages.

Diagnostics

The Diagnostics tab shows visualizations for:

• Model Assessment visualization, which you can break down on Actual vs. Predicted or Residual conversions.

  To break down the visualization, select Actual vs. Predicted or Residuals from the Breakdown list.

• Model fitting metrics table, showing the following columns for each conversion metric:

  • Actual Conversion

  • Modeled Conversion

  • Residual Conversion (difference between actual and modeled conversion)

  • Model quality score values:

    • R2 (R-squared), which tells how well the data fits the regression model (the goodness of fit).

    • MAPE (Mean Absolute Percentage Error), which is one of the most commonly used KPIs to measure forecast accuracy and expresses the forecast error as a percentage of the actual value.

    • RMSE (Root Mean Square Error): which shows the average error, weighted according to the square of the error.

  To download a CSV file containing the data for the table, select .

• Touchpoint effectiveness table, representing the outcome of the Attribution AI algorithmic model. The data for this table is only generated for specific periods of time. Select As of xx/xx/xx, xx:xx TZ for more details.

  The visualization shows, in descending order of Efficiency measure , for each touchpoint:

  • Paths touched: visualizes the percentage of paths achieving conversion and percentage of paths not achieving conversion. For a touchpoint, you see more attributed conversions when the attribution conversion ratio is high. This ratio compares the percentage of paths that lead to conversion versus the percentage of paths that do not lead to conversion.
  • Efficiency measure: generated by the algorithmic attribution model, the efficiency measure indicates the relative importance of a touchpoint toward conversion, independent of touchpoint volume. The efficiency is measured on a scale of 1 to 5. Note that higher touchpoint volume does not guarantee higher efficiency measure.
  • Total volume: The aggregate number of times a user touches a touchpoint. The number is inclusive of touchpoints that appear on a path achieving conversion as well as paths not resulting in conversion.

Model drift detection

If model drift is detected, you see a Model drift detected notification at the top.

Select Hide to hide the notification. The notification will re-appear the next day or at the next login.

Historical overview

The Historical overview tab shows visualizations for:

Conversion and Spend by Fiscal Qtr and Product

This visualization represents the conversion and spend distribution across various quarters within the given date range. The visualization helps to identify high-performing quarters where spend is driving conversions.

Spend by Channel

This visualization represents the spend distribution across various channels within the given date range. The visualization supports quick identification of which channels receive the most spend.

Touchpoint Spend

This visualization represents the spend distribution across paid touchpoints for each quarter within the given date range. The visualization enables understanding of which touchpoints are prioritized within specific channels and quarters. The visualization helps to identify channel spending patterns and trends, particularly channels with low and infrequent spending over time.

To can select an alternative spend-based channel to display for this visualization:

• Select a channel from Channels.

Touchpoint Volume

This visualization represents the volume distribution across all touchpoints for each quarter within the given date range.

To can select an alternative volume-based channel to display for this visualization:

• Select a channel from Channels.

Edit

You can edit the name, description and the scheduling of training and scoring of the model.

1. Select Edit

2. In the Edit model dialog:

   • Enter a new Name and Description.

   • To enable scheduling, enable Status. You can only enable scheduling for models that are trained and scored.

     1. Select a Scoring frequency:

        • Daily: Enter a valid time (for example 05:22 pm) or use .
        • Weekly: Select a day of the week and enter a valid time (for example 05:22 pm) or use .
        • Monthly: Select a day of the month from the Run on every dropdown menu and enter a valid time (for example 05:22 pm) or use .

     2. Select a Training frequency from the dropdown menu: Monthly, Quarterly, Yearly, or None.

     

3. Select Save.