Algorithmic modeling in AudienceManager can help you analyze a subset of your site visitors and target others who are similar.  If you’re just getting started with AudienceManager or even if you just need to brush up on this feature, here are some tips that should help you get the most out of the tool.

1. Build a granular trait taxonomy

Imagine someone gives you a task to compile a list of  characteristics that make Adobe Digital Marketing Summit attendees unique from the general population.  At the top of the list you might add the really obvious distinguishing characteristics, like:

  • Works in digital advertising
  • Interested in audience targeting

A little further down you might list some looser defining factors:

  • Has a smartphone
  • Has thick glasses frames
Adobe Summit attendees

We know that Summit attendees are unique, but what is it exactly that makes them unique?

When you build a model from a baseline, you’re asking AudienceManager to perform the same task:  compile a list of traits that make the baseline unique from the general population.  The model results will be more accurate when there are more traits, or defining factors, to analyze.  Think of your traits as tools in your tool shed; the more tools you have, the better the finished product will be.

Note:  the modeling methodology above is simplified for illustration.

2. Select a good baseline

  • The baseline should represent a small subset of your visitors. Remember, the goal is to take a group of visitors who exhibited a desirable trait and determine what makes them unique (so you can expand your targeting to users who look like the baseline). If your baseline already represents a majority of users, then there is no way to determine how they are unique from the general population since they are already the general population.
  • Make sure the baseline isn’t TOO small.  If the audience size is insignificant there will be little data to look at when determining the model results.  To illustrate, if you have 10 users in your baseline and each user has a completely different trait profile (or only a few traits that differ from the general population), the model will not produce results because there is no real difference between the baseline and the general population.
  • The baseline trait or segment shouldn’t be newly created.  A general guideline is to allow for at least 1-2 weeks of data collection on a baseline trait/segment before creating a model.  30 days or more is ideal.
    • Note: AudienceManager will periodically re-run the model so even if your baseline doesn’t have much history to start, it will re-analyze user profiles to update the model.

3. Select a short look back period

When you select a short look back period (say, 30 days), you’re telling AudienceManager “I’m only interested in what the baseline looked like over the last 30 days.”  Remember that AudienceManager will re-run the model periodically so in this case, each time we do, your model will always represent the most recent 30 day snapshot of the baseline.

Lookback period

Using a short look-back period ensures each time the model is run, it will use the most recent set of data

4. Create an “exclude” data source

As part of your trait taxonomy, you probably already have “site visitor” traits which represent a majority (or all) of your users.  If this trait is considered when running the model, you’ll get a result that looks like a flat line and you won’t be able to scale the algorithmic trait up or down.  To avoid, you’ll want to classify your site visitor traits as a separate data source.  This also goes for very large or generic traits, like “Operating System” or “Language”.

flat line chart

Including large traits like “Site Visitor” traits in your model will produce a flat line.

Ask your AudienceManager consultant about setting up the appropriate “exclusion” data source(s) for your trait taxonomy.  Once set up, you can change a trait’s data source in the Trait Edit screen.

When you have your trait taxonomy grouped into your various data sources, you can leave a “site visitor” (or similar) data source deselected when creating a new model, under “Select Available Data.”

algorithmic curve chart

When you exclude generic traits like “Site Visitor” traits, the model should produce a nice algorithmic curve.

5. Apply logic

Most of all, think through the business case and make sure your model makes sense.  Creating a model based on visitors who have hit the homepage of your site probably doesn’t satisfy any use case, while creating a model based on specific product purchasers probably does.  Make sure your baseline answers the question “who are my most valuable visitors” and you’ll be off to a good start.