Build models

Enter your model Name, for example Demo model. Enter a Description, for example Demo model to explore AI features of Mix Modeler.

Select Next to continue to the next step. Select Cancel to cancel the model configuration.

In the Conversion goal section:

1. Select a conversion from the Conversion dropdown menu. The available conversions are the conversion that you defined as part of Conversions in Harmonized datasets. For example, Online Conversion.

2. You can select Create a conversion to create a conversion directly from within the model configuration.

In the Marketing touchpoints section, you can select one or more marketing touchpoints, corresponding to the marketing touchpoints you defined as part of Marketing touchpoints in Harmonized datasets.

1. Select one or more marketing touchpoint from the Touchpoint include dropdown menu.

   • You can use to remove a touchpoint.
   • You can use Clear all to remove all touchpoints.

2. You can select Create a touchpoint to create a marketing touchpoint directly from within the model configuration.

By default, a score is generated for all the data in your harmonized view. To only score a subset of the population, define one or more filters using containers in the Eligible data population section.

• For each container, define one or more events.

  1. For each event:

     1. Select a metric or dimension from Select harmonized field.

     2. Select the appropriate operator: equals, not equals, less than, greater than, starts with, doesn’t start with, ends with, doesn’t end with, contains, doesn’t contain, is in, or is not in.

     3. Enter or select a value at Enter or select value.

  2. To add an additional event in the container, select Add event.

  3. To remove an event from the container, select .

  4. To filter using all or any of multiple events defined in the container, select Any of or All of. The label correspondingly changes from Include … Or … to Include … And ….

• To add an eligible data population container, select Add eligible population.

• To remove an eligible data population container, within the container, select , and select Remove container from the context menu.

• Select And and Or between containers to build more complex definitions for your eligible data population.

To add datasets containing external factors to your model, use one or more containers in the External factors dataset section. An example of external factors are S&P indices.

• For each container:

  1. Enter a External factor name, for example External Factors.

  2. Select a dataset from the Dataset dropdown menu. You can select to manage datasets. See Datasets for more information.

  3. Select an option from the Impact on conversion dropdown menu: Auto select, Positive or Negative. The default option is Auto select, which allows the model to determine the impact. You can override the default.

• To add an additional external factors dataset container, select Add external factor.

• To remove an external factors dataset container, select .

To add datasets containing internal factors to your model, use one or more containers in the Internal factors dataset section. An example of internal factors are email marketing data.

• For each container:

  1. Enter a Internal factor name, for example Email Marketing Data.

  2. Select a dataset from Select a dataset. You can select to manage datasets. See Datasets for more information.

  3. Select an option from the Impact on conversion dropdown menu: Auto select, Positive or Negative.

• To add an additional internal factors dataset container, select Add internal factor.

• To remove an internal factors dataset container, select .

To define the lookback window for the model, enter a value between 1 and 52 in Give contribution credit to touchpoints occurring within … weeks prior to the conversion.

Select Next to continue to the next step. If more configuration is needed, a red outline and text explains what additional configuration is required.
Select Back to go back to the previous step.
Select Cancel to cancel the model configuration.

In the Spend share section:

• To use historical marketing investment ratios to inform the model when marketing data is sparse, activate Allow spend share. This setting is recommended, especially in the following scenarios:

  • A channel doesn’t have enough observations (for example, low frequency of spend, impressions or clicks).
  • You are modeling spiky but regular, and potentially high-spend media (like TV for some brands), where data may be sparse.

  note note
  NOTE
  For one-off investments (for example a Super Bowl ad), consider to incorporate that data as a factor rather than to rely on spend share.

In the MTA enabled section:

• To enable MTA features for the model, activate MTA enabled. If you have enabled MTA, multi-touch attribution insights are available after you have trained and scored your model. See the Attribution tab in Model insights.

In the Prior knowledge section:

1. Select the Rule type, which is by default Absolute values.

2. Specify contribution percentages for any of the channels listed under Name, using the Contribution proportion column.

3. Where appropriate, you can add for each channel a Level of confidence percentage.

4. When needed, use Clear all to clear all input values for the Contribution proportion and Level of confidence columns.

Turn on Enable scheduled model scoring and training.

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 .

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