Creating audience look-alikes

This guide offers step-by-step instructions on creating a look-alike audience by leveraging predictive variables.

Note: A list of potentially predictive variables predefined by your system administrator is a necessary prerequisite for using look-alike audiences. See Selecting variables for audience look-alikes.

Creating a new ‘Find Look-alikes’ solution

To create a new look-alikes solution:

1. Begin with the audience you want to identify look-alikes for.

2. Click Add New, to add a new solution.

   

3. Add a Find Look-Alike solution to your workbook and give it a name.

4. Click Run to initialise your Find Look-alikes solution.

   

5. Select a comparison audience, if required.

   Note: A comparison audience is the backdrop that you wish to compare against.

6. Choose the predictive Variables to use for analysis.

   Note: You need to select at least two predictive variables.

7. Click Run to start analysing characteristics in your data.

   The analysis runs in the background, notifying you when complete.

Refining your find look-alikes solution

At this point, Orbit has now run a profile using your choices. The dimensions have been selected and ranked for the insight they provide using Incremental Insight.

Note: Incremental Insight effectively reduces the influence of highly correlated variables. However, once variables are included in the model, they are scored using their original ‘undampened’ weights, not the adjusted weights from Incremental Insight. This approach leads to multiple overly dominant correlated variables in the model, as they add similar signals repeatedly with full weight. We are working on resolving this issue.

Note: Incremental Insight is a data selection method geared towards optimising the choice of predictive dimensions in modelling. Unlike traditional methods that focus solely on the standalone predictive power of each dimension, Incremental Insight factors in mutual correlations.

The incremental insight principle is straightforward:

• Begin with the dimension having the highest predictive strength

• For each subsequent choice, account for its similarity (correlation) with dimensions already chosen

Note: In some circumstances you may have more records in an audience look-alike than you have positive scores. These circumstances may result in model scores with less than zero being included in your results.

This approach ensures a balance between the raw power of individual dimensions and the diversity that they bring, preventing redundant or overlapping information from clouding the model.

You can see the most predictive variables ranked here:

If you wish to add or remove variables from the calculation, or change the comparison audience, the settings are accessible here.

Note: Creating look-alikes using Incremental Insight requires that the best variables are chosen, so preceding variables in the list are selected automatically.

Creating a new look-alikes audience

Now you can create a new audience workbook which contains records that have similar characteristics to the ones in your look-alike solution.

1. Pick the most insightful variables and then choose **Find Look-alikes**.

   

2. Choose the settings to create your new audience of potential prospects. You choose how many more records you want, and the new audience includes this number of the best scoring (most alike) records for you to market to.

   

3. Click Create to generate a new look-alike audience with these settings. The process runs in the background, allowing you to perform other tasks.

4. A notification pops up when your audience is ready. This might take a few moments to appear.

   

   Note: The original audience is locked to preserve the settings.

   

See Using audience look-alikes.