On one project I reviewed, the marketing team had five personalization variants targeting "Personal loan interested visitors." Each variant defined that group slightly differently. One checked an affinity, one checked recent page views, two checked both with different time windows, and one had been copied from a test and never cleaned up. Nobody could say which was right, because there was no single definition to point at.

Audience Builder in SitecoreAI solves this by letting you define a visitor group once and reuse it. This guide walks through the full workflow step by step: create, define, estimate, inspect, set live, and verify membership.

Before you begin

Audience and Insights is being rolled out in phases, so confirm Audiences is available under Optimization in the environment you're working in.

Prerequisites

You need:

• Access to Optimization in a SitecoreAI environment with Audience and Insights enabled.
• Reliable VIEW and IDENTITY events, plus identity rules and affinities configured. My earlier guides cover both.
• A written business definition of the audience you want to build, agreed with whoever will use it.

That last item matters more than any UI setting.

Why use Audience Builder

An audience is not a saved filter. It's a reusable definition that SitecoreAI evaluates against profile data and, once live, records as membership on matching profiles.

That gives you one place to define "returning customers" or "EV-interested visitors," and one place to fix it when the definition changes.

How it works

The key distinction is between Draft and Live. In Draft, you're editing and refining rules. Once Live, SitecoreAI evaluates profile data against those rules and saves membership on profiles that match.

The step developers usually misjudge is the last one. Membership is evaluated several times a day, not instantly. If you set an audience live and immediately check your test profile, it probably won't be there yet.

Step 1: Create the audience

1. In SitecoreAI, open Optimization.
2. Select Audiences.
3. Click Create audience.
   
   
   
   
   
4. Enter a descriptive name.
   
   
   
   
   
   
   
5. Click Create.
   
   

The audience opens in Audience Builder with Draft status.

Name it for its business purpose:

• Returning Customers
• Campaign Visitors
• Interested in Electric Vehicles
• High-Intent Product Visitors

Avoid names like Audience1 or Rule_Test_Final. The name is often the only documentation other teams see.

Step 2: Define the audience in business terms

Before adding a rule, write one sentence describing who belongs in the audience. For example:

Visitors who showed interest in suv vehicles and recently viewed product content.

Then translate that sentence into conditions. I do this on paper (or in the ticket) before opening the builder. It surfaces gaps early, such as "what counts as recently?" or "do we have an affinity for suv?", while they're still cheap to resolve.

Step 3: Add audience rules

Audience Builder organizes data into attribute groups:

• Profile
• Sessions
• Events
• Traits

An attribute is a field from one of these groups. A rule is the condition a profile must satisfy.

1. Open the draft audience.
2. Select an attribute group.
3. Choose the attribute.
4. Set the comparator and value.
   
   
    
   
   
5. Add further rules or rule groups as your business definition requires.

A simple example combining three groups:

Language = EN 

AND Last Seen At is in the past 2 days 

AND Page View Event is = Sedan

The attributes you see depend on the data your environment actually collects. If an attribute you expected is missing, that's usually a data-collection gap rather than an Audience Builder issue. Check the profile first.

Keep related conditions together. Another developer should be able to read the audience without reverse-engineering the logic.

Step 4: Estimate the audience size

Never set an audience live without this step.

1. Open the audience and review the rules.
2. In the right pane, locate Audience size.
3. Click the calculate icon.
   
   
   

4. Wait for the calculation to finish. It can take a few minutes.
5. Review Estimated profiles.

Treat the number as a sanity check against your business definition. If "High-Intent Product Visitors" comes back larger than your total monthly traffic, something in the logic is too broad. The opposite also happens: an estimate of zero often means a value doesn't match the underlying data, such as Electric in the rule and electric in the affinity taxonomy.

I like to write down the expected rough size before calculating. It forces you to have an opinion, which makes a wrong result much easier to spot.

Step 5: Edit, reorder, or delete rules

While the audience is in Draft, you can change it freely.

To edit a rule:

1. Open the audience.
2. Locate the rule.
3. Change the attribute, comparator, or value.
4. Save.
5. Recalculate the estimate.

To delete a rule, remove it and recalculate.

To reorder rules or rule groups, arrange them so related conditions sit together. Order helps people understand and maintain the definition, especially once you have several rule groups.

To delete an audience, follow your organization's process. An audience other teams rely on shouldn't disappear without notice.

Step 6: Inspect the definition

The visual builder is easy to read, but it isn't the whole story. Before going live, check what SitecoreAI actually generated.

1. Open the audience.
2. Click the ellipsis in the top-right of Audience Builder.
3. Select View code.
   
   
   

SitecoreAI shows generated code and a JSON representation of the audience. Use this to confirm the rule structure, especially how rule groups relate to each other. Inspect mode is also available for examining the configuration and its rule groups, and it's the first thing I open when troubleshooting a complex audience.

My review sequence before go-live:

Each layer catches a different mistake. The generated definition catches logic that looks right visually
but groups differently than intended.

Step 7: Set the audience live

1. Open the draft audience.
2. Confirm all rules are complete.
3. Review the estimated size one last time.
4. Click Set live.
5. Change Schedule time when the audiance will update. You can select 6,8,12 or 24 hours.
6. Click Apply  

   
   

   

The status changes from Draft to Live. SitecoreAI calculates an initial estimate and then evaluates profiles against the rules several times during the day, saving membership on matches.

If you need to change a live audience, pause it first. Don't treat a live audience like a draft. After making changes, re-check the estimate and set it live again.

Step 8: Verify membership on a profile

Going live isn't the finish line. Confirm the audience is actually matching the profiles you expect.

1. Open Optimization > Profiles.
2. Find a profile that should match.
3. Open the profile details.
4. Use Inspect to review audience membership.
   
   
   

Allow time for evaluation before checking. If membership is missing, compare the profile's underlying data with the rules, not the other way around.

Troubleshooting

The audience is far too large

Check whether:

1. Rules use the intended attribute group.
2. Conditions are broader than the business definition.
3. Rule groups have the logical relationship you intended. Confirm this in View code.
4. Selected values are correct.

An expected profile isn't a member

Start with the profile, not the audience:

1. Is the required event present?
2. Is the required profile attribute populated?
3. Is the expected affinity there?
4. Does the profile satisfy every rule?
5. Is the audience actually live, and has evaluation had time to run?

Changes aren't appearing

Check the lifecycle state. A live audience must be paused before you modify it. After changes, validate the estimate and set it live again.

Common mistakes

Going live without checking the estimate. A technically valid audience can still describe the wrong people.

Building on an unstable taxonomy. If affinity or event values are inconsistent, the audience rules become impossible to maintain.

Editing a live audience directly. Pause first.

Expecting instant membership. Evaluation runs several times a day. Set expectations with the business before launch.

Duplicating audiences. If the same group is needed in several places, build one audience and reuse it.

Real project tips

Treat audiences as governed configuration, not ad-hoc segments. For each one, record:

• Business purpose
• Owner
• Data dependencies
• Rule logic
• Expected approximate size
• Downstream consumers
• Review cadence

A simple shared table is enough. When an affinity value gets renamed or an event changes, the "data dependencies" column tells you exactly which audiences to retest.

Summary

Audience Builder turns profile data into reusable visitor groups, but the tool is the easy part. The real work is making sure every audience maps to a clear business requirement, depends on reliable data, and gets validated before it goes live. The workflow I follow every time:

Create → Define → Estimate → Inspect → Set live → Verify membership

When that becomes routine, audiences stay a governed part of your SitecoreAI architecture instead of a growing list of segments nobody trusts.

References

Audiences | Sitecore Documentation