Top SFMC Segmentation Approaches in 2026

‍

Segmentation in Salesforce Marketing Cloud has never had a single "right" method. Teams mix several, and the mix usually reflects who is available to build the audience rather than what the campaign needs.

This guide compares the main approaches in use in 2026, what each does well, and where each one starts to strain.

‍

The goal is not to crown a winner. It is to help you match the method to the job.

‍

Point-and-click segmentation

‍

Filters and Audience Builder

The most accessible option is the visual tooling inside SFMC. Filtered Data Extensions and Audience Builder let a marketer define rules through the interface, with no code.

For simple cases, such as "subscribers in a given country who joined this year", this is fast and easy to review.

‍

Where it strains

Visual filters work best on a single, well-structured data source. Once an audience depends on engagement history, purchase data and exclusions at the same time, the rule logic gets hard to read and harder to maintain.

Many teams reach that point sooner than they expect, and then hand the request to a technical colleague.

‍

A practical test is to count the sources an audience touches. If the answer is more than one, visual filters probably stop being the simplest option.

Another warning sign is review time. When a colleague cannot tell at a glance what a segment contains, the rules have outgrown the interface.

‍

None of this is a criticism of the visual tools. They are well suited to the narrow set of cases they were designed for.

‍

SQL Query Activities in Automation Studio

‍

Why SQL remains the workhorse

SQL Query Activities are still the most flexible way to build audiences. You can join data extensions, reach the system data views, and write the result into a destination data extension on a schedule.

Salesforce's Trailhead material on SQL for Marketing Cloud Engagement covers the core commands: SELECT, FROM, JOIN and WHERE.

‍

The documented limits to plan around

Salesforce documents hard limits that shape how you design queries. A single query times out after 30 minutes, and Salesforce recommends keeping queries under 5 minutes for best performance.

Soft limits also apply, including a recommended maximum of 4 data extensions and 3 joins per query, and daily query volumes that depend on your edition.

‍

None of this makes SQL a poor choice. It means query design is a skill, and the skill sits with a small number of people.

In practice, this is why query reviews matter. A query that works on a small test set can behave very differently on a large production table.

‍

Selecting only the fields you need, rather than every column, is a simple habit that helps queries finish well inside the time limit.

Scheduling matters too. Because results land in a destination data extension, a well-built query keeps an audience fresh without anyone rebuilding it by hand.

‍

Engagement-based and predictive signals

‍

Using engagement data as the rule

Many of the most useful audiences are defined by behaviour: who opened, who clicked, who went quiet. In SFMC, that history lives in data views such as _Sent, _Open and _Click.

Salesforce also offers Einstein Engagement Scoring, which scores contacts on their likelihood to engage, and these scores can feed audience rules.

‍

The caveat on predictive scores

A score is an input, not an audience. You still need a clear rule for what to do with it, and a way to combine it with exclusions and business logic.

Treat predictive signals as one filter among several.

‍

For example, a team might combine a low engagement score with a recent purchase, and exclude anyone already in an active journey. Each condition is simple, but the combination is where the value sits.

That combination is exactly the kind of logic that tends to end up in SQL.

‍

Data Cloud segmentation

‍

When a unified profile is the foundation

Teams that have adopted Data Cloud can build segments on unified profiles drawn from many sources. This is attractive for organisations whose customer data lives across several systems.

It is a larger architectural decision than choosing a segmentation method, so it usually sits within a wider data strategy.

‍

How it coexists with Engagement-level work

Even with Data Cloud in place, many teams keep building campaign-level audiences inside Marketing Cloud Engagement. The two approaches often run side by side.

The practical question is where the audience needs to be activated. If the send happens in Email Studio or Journey Builder, the data eventually has to be available there in a usable form.

‍

Because of this, Data Cloud adoption rarely removes the need for query skills. It changes where some of that work happens.

‍

Natural-language and AI-assisted segmentation

‍

What changes for the marketer

The newer approach is to describe the audience in plain language and let an assistant produce the query. Instead of writing a join, a marketer might ask: "Contacts who bought in the last 90 days but have not opened an email in 60."

The output is still SQL, so it stays inspectable and compatible with the automations teams already run.

‍

That matters because SQL is a shared language. A reviewer who knows the data views can read the result, correct it, and approve it before anything is sent.

It also means teams do not have to abandon existing automations to try the approach.

‍

What to check before relying on it

Ask whether the generated query is visible, whether it respects the data views and limits above, and whether a technical colleague can review it.

AI assistance removes the typing, not the need for sound logic.

‍

It also helps to keep a human in the loop for sensitive audiences, such as suppression lists, where a wrong rule has real consequences.

Used this way, natural language becomes a faster first draft, and expert review remains the safeguard.

‍

How to choose between them

‍

Match the method to the audience

• Simple, single-source rules: visual filters or Audience Builder
• Multi-table logic and scheduled refreshes: SQL Query Activities
• Behaviour at scale: data views, with or without engagement scores
• Cross-system profiles: Data Cloud
• Fast requests from non-technical teams: natural-language generation

‍

Match the method to the team

The most useful question is often organisational: who can build and maintain this audience when the author is away?

Approaches that depend on one specialist create a bottleneck, whatever their technical merit.

‍

Documentation and naming conventions help whichever method you pick. A clear name and a one-line description on each audience save hours later.

Finally, revisit the choice regularly. A method that suited a team of two may not suit a team of twenty running campaigns across several markets.

‍

See QAiry in action

‍

QAiry is a conversational assistant for Salesforce Marketing Cloud that turns plain-language requests into SQL you can review.

You can see how it works at qairy.com/product-demos, or try it yourself at qairy.com/try-it-free.