Build a Best Fit Attribution Model


Read more about Data Workbench’s End-of-life announcement.

Open Best Fit Attribution from the Premium menu and follow these steps to build a Best Fit Attribution model.

See an overview of Best Fit Attribution.

  1. Open Best Fit Attribution.

    Open a workspace and click Premium > Best Fit Attribution.


    Best Fit Attribution is an Adobe Analytics Premium feature that requires you to enable Premium in your Profile. It requires you to update your certificate and add the Premium profile to your profile.cfg file. See DWB Server upgrade: 6.2 to 6.3 for DWB 6.3.

  2. Set the Success metric.


    You can either drag a metric from a Finder table to the left pane of the Attribution visualization, or select from the Inputs menu.

    Click Inputs > Set Success. The metric menu will open.

    Select a metric that identifies a successful conversion.

  3. (optional) Set the Revenue metric.

    Set a metric to evaluate revenue across the conversion process.

  4. Set the Touch metric.


    Setting a Touch Metric is only required if you are trying to build Success metrics automatically by dragging dimension elements onto the visualization.

    Click the Inputs menu and select Set Touch, or drag a metric from the Finder.

    This will be used to derive channel metrics when dimension elements are used as inputs.

  5. Set a Success window.

    Click Inputs > Success Window. Select a date range from a table and then name the Success window. Click Workspace Selection and the selected dates will be assigned as the range of time for the Success metric.


    Since the Success window is a workstation selection, you can include any dimension(s) to your Success window.

  6. Set a Touch Window.

    Click Inputs > Touch Window. Select a date range from a table and then name the Touch window. Click Workspace Selection and the selected dates will be assigned as the range of time for the Success metric.

    By default, the Touch window will be set to the same time period as the Success window.

  7. (optional) Set a Training Filter.

    You can also specify a Training Filter in the workspace to filter visitor data.


    In setting both the Success and Touch windows, you can apply the Training filter to the current workspace selections to further limit your data.


    The training set is always drawn from visitors who satisfy the Success window. By filtering using the Filter Editor, you can create a subset of visitors reported in the Success window.

  8. Specify channel metrics that represent touches.

    Either drag metrics to the visualization, or choose them from the Inputs > Add Channel menu. If you do not already have metrics defined for campaigns or channels, but do have dimensions representing channels, the visualization can build them for you automatically with the specification of a Touch metric.

    For example, with the Touch metric set to Hits, and given a dimension called Media Type with elements that include things like Email, Press Release, Print Ad, and Social Media, the visualization will generate Channel metrics of the form Hits where Media Type = Email when you drag and drop the element(s) onto the visualization.

  9. Press Go.

    The Best Fit Analysis process will run, and a chart will display attributions per channel based on the selected inputs.


    Right-click Model Complete on the completed analysis to see statistics for the attribution model.

When complete, a graph will display an attribution model calculated per channel, and a distribution of the Revenue metric (if set). The model can be saved internally or exported to other systems.


Streaming, Online and Offline modes produce different effects when building an attribution model based on the latency of the data being evaluated. In Streaming mode, the detail Model Complete message will display. In Online and Offline modes, the detail Local Model Complete will display.

Options menu

The Options menu provides advanced features to set up and display Best Fit Attribution analysis.

Options menu Description
Set Training Filter The Training Filter is used with the Success Window to filter the population when building the attribution model. This will provide a subset of data that includes only the visitors that you want to analyze.

Note: Experienced users can also leverage the flexibility of filters to focus beyond the time line of the Success and Touch Windows. For example, in addition to selecting a time range, you can select a set of Referring Domains to only examine the attribution for users from those domains.

Show Complex Filter Description Displays the filter code for the Training Filter, Success Window, and Touch Window.
Save Model Saves the current attribution model for future use.
Load Model Opens a previously saved attribution model.
Presentation View Hides the top menu bar for presentation.

Options > Advanced includes features to set the training set size and specify the approach to take in the case of a class imbalance.

Advanced > Training Set Size

Sets the training set size.

Note: The default training size is Large for 250,000 visitors.

  • Tiny = 50,000
  • Small = 75,000
  • Normal = 100,000
  • Large = 250,00
  • Huge = 500,000
Advanced > Class Balance

Identifies and defines the number of input records to generate for a class imbalance issue based on dataset size.

Reset and Remove options Description
Reset Model From the Reset menu, select Reset Model to clear the visualization but keep input metrics.
Reset All From the Reset menu, select Reset All to clear the visualization and the input metrics.
Remove Right-click on any input and select Remove to clear the metric from the selected input.
Remove All Right-click on Channels and select Remove All to clear all input metrics.

On this page