Planning a Predictive Goal

For your predictive goals to be as effective as they can be, there are a few things to address before you build them.

# In this article

# Understand your objective

When you create a predictive goal, it's important to be clear about what you want to learn from it. Questions you might consider include:

• Do you want to discover whether users are performing specific tasks, or general tasks?

  For example, maybe you want to find users who are likely to make a specific type of purchase in the next month (like the purchase of a discounted item), or maybe you're interested in more general actions (like finding users who'd make any kind of purchase).

• Are you looking for potential trends?

  Because predictive goals don't have finite start and stop dates (they "roll" from one refresh to the next), a single goal can tell you how many users are likely to perform an action for a certain period of time, starting when you create a goal, and for the same period of time after subsequent weekly refreshes. This can help you track, for example, how many users might upgrade to a premium account in a one month period.

• Do you want to evaluate brand engagement based on unique custom events?

  If you're already familiar with Brand Affinity™, you can think of this as the next step in measuring users' loyalty to your brand. Instead of assessing engagement based on historical clicks and opens, use Predictive Goals to see who's likely to engage based on your unique events.

  For example, if building a sense of community is important to your brand, you might be particularly interested to discover who's likely to perform actions represented by custom events like BlogViewed or ReferFriend.

# Verify your data for Predictive Goals

Before you build a predictive goal, verify that you have enough and the right type of data.

# Amount of data

At a minimum, your Iterable project should have:

• Six months of historical data
• 100,000 unique users
• 100,000 event and user profile field occurrences across users, with at least 20—30 active events that have been frequently updated across all users in the past month.

# Type of data

The criteria for a predictive goal can include user profile fields, system events, custom events, and purchase events.

• Use user profile fields to track information that defines a user, in some way. For example, you could use them to query the types of accounts users have.

• Use Iterable system events and your unique custom events, to track actions users take. For example, you could query when users click an email, or when they perform an action that's specific to your business (maybe a travel site would track booking types, with custom events like BookingHotel, BookingAir, and BookingAuto).

• Use Iterable Purchase events to query a dollar value and the number of purchase occurrences.

For more information about how Predictive Goals uses the data you select, talk with your customer success manager.

# Select meaningful data points

Once you've verified that you have the right amount and type of data for effective predictions, consider how your data maps to the business goal you want to track.

# Example goal: Reduce churn

To reduce the number of users who might stop using your product or service, create a goal that allows you to segment users by how likely they are to churn, so you can send customized messaging to each segment.

People who are actively engaged with your service are most likely to support your goal of reducing churn, so creating a goal that finds them is a good place to start.

• If your project tracks user engagement through user profile information (like account status–either active, inactive, or churned), your goal criteria might include a contact property like this:

  AccountStatus = active in a given 1 month

• Or, if your project tracks this type of information through actions users perform, your goal criteria might include a custom event like this:

  AccountSignUp = 1 in a given 1 month

Your approach depends on how your data’s set up and how you want to use it, but either path helps you find the audiences who are most and least likely to support your goal–reducing churn.

• Send messaging that rewards users for their loyalty when they are more likely to support your goal (less likely to churn), so they continue supporting you.

• Send messaging that reengages users who are less likely to support your goal (more likely to churn) to help move them back into the group of ideal users–those who are unlikely to churn.

# Example goal: Increase subscriptions

If you’re focused on increasing the number of subscribers to your service, and your project tracks users who subscribe, you might find people who are likely to subscribe in the future with goal criteria like this:

• If a contact property (user profile field) tracks users who subscribe:

  Subscribed = true in a given 1 month

  NOTE

  The contact property Subscribed = true includes users in segmentation who've already subscribed. If your messaging will encourage users to subscribe, exclude them from the selected segment, so they don't receive messaging that encourages an action they've already taken. See Exclude users from segmentation.

• If a custom event is triggered when a user clicks a Subscribe Now button from their trial:

  ClickSubscribe ≥ 1 in a given 1 month

When your prediction is ready, select the users with the highest likelihood of conversion and send them a limited-time offer to encourage them to upgrade soon.

# Example goal: Increase app downloads

If you're looking to increase the number of users who download your app, and your project tracks app downloads, you could find users who are likely to download your app in the next month like this:

• If a contact property (user profile field) tracks users who have your mobile app:

  HasMobileApp = true in a given 1 month

• If, instead, a custom event is triggered when a user clicks a Download App button:

  DownloadApp ≥ 1 in a given 1 month

With either criteria, choose two segments for messaging designed to help increase downloads of your app.

• Users who are more likely to download can receive a promotional offer that encourages them to install within the next week so you keep them on the trajectory they're already on.

• Users who are less likely to download can receive messaging that considers why they’re less likely to do so. For example, consider whether these users aren’t aware of the benefits of using the app, or perhaps they have an alternate preferred channel (maybe they don’t use mobile).

  To address both scenarios, you could send this segment a campaign that highlights the benefits of the app, but also provides an option to specify their preferred method of communication.

# Next steps

• Learn how to create a new predictive goal. See Building a predictive goal.

• See how a fictional company built predictive goals to meet their needs, check out Use case: Reward loyalty with memorable experiences.